Skip to content

Listings

All listings: OpenRouter

Every listing for the models on page 2 of 26 of OpenRouter, including each model's other versions and routes. Back to OpenRouter

Sold by OpenRouter

229 listings on this page sold by OpenRouter.

Select a score for details.

About these columns
llminfo IQ
llminfo IQ (capability index). How llminfo IQ is computed
Coding IQ
llminfo Coding IQ (coding capability, its own scale). How Coding IQ is computed
Value
Value for money. 50 means as expected for its results. Above 50, cheaper per task. How Value for money works
Yap
Tokens used per coding task vs. models with the same results. Includes agents that write code. Chattiness on maths doesn't predict it. How the Yap score works
Modelllminfo IQCoding IQInput $/1MOutput $/1MMax Input TokensValueYapEpoch Capabilities Index
Gemini 3.8 Flash
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
139.6
From 325 benchmarks. Plenty of data behind it.
121 ±8
Coding IQ 121.0, 95% interval 113–129. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.825cache read $0.0825$4.1251M
55.0
Good value (55.0). Cheaper per task than its benchmark results predict. Plenty of data behind it.
64 ±10
Chatty (64 ±10). Generates more tokens than models with similar coding results. Measured at high effort. A fair amount of data behind it.
158.8
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Gemini 3.8 Flash
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
google/gemini-3.8-flash:batch · batch
139.6
From 325 benchmarks. Plenty of data behind it.
121 ±8
Coding IQ 121.0, 95% interval 113–129. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.375cache read $0.0375$1.8751M
55.0
Good value (55.0). Cheaper per task than its benchmark results predict. Plenty of data behind it.
64 ±10
Chatty (64 ±10). Generates more tokens than models with similar coding results. Measured at high effort. A fair amount of data behind it.
158.8
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Gemini 3.8 Flash
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Inference by Google
139.6
From 325 benchmarks. Plenty of data behind it.
121 ±8
Coding IQ 121.0, 95% interval 113–129. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.75cache read $0.075$3.751M
55.0
Good value (55.0). Cheaper per task than its benchmark results predict. Plenty of data behind it.
64 ±10
Chatty (64 ±10). Generates more tokens than models with similar coding results. Measured at high effort. A fair amount of data behind it.
158.8
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Gemini 3.8 Flash
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
google/gemini-3.8-flash:batch · batch+global
Inference by Google
139.6
From 325 benchmarks. Plenty of data behind it.
121 ±8
Coding IQ 121.0, 95% interval 113–129. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.375cache read $0.0375$1.8751M
55.0
Good value (55.0). Cheaper per task than its benchmark results predict. Plenty of data behind it.
64 ±10
Chatty (64 ±10). Generates more tokens than models with similar coding results. Measured at high effort. A fair amount of data behind it.
158.8
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Gemini 3.8 Flash
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
flex
Inference by Google
139.6
From 325 benchmarks. Plenty of data behind it.
121 ±8
Coding IQ 121.0, 95% interval 113–129. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.375cache read $0.0375$1.8751M
55.0
Good value (55.0). Cheaper per task than its benchmark results predict. Plenty of data behind it.
64 ±10
Chatty (64 ±10). Generates more tokens than models with similar coding results. Measured at high effort. A fair amount of data behind it.
158.8
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Gemini 3.8 Flash
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
global
Inference by Google
139.6
From 325 benchmarks. Plenty of data behind it.
121 ±8
Coding IQ 121.0, 95% interval 113–129. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.75cache read $0.075$3.751M
55.0
Good value (55.0). Cheaper per task than its benchmark results predict. Plenty of data behind it.
64 ±10
Chatty (64 ±10). Generates more tokens than models with similar coding results. Measured at high effort. A fair amount of data behind it.
158.8
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Gemini 3.8 Flash
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
global/flex
Inference by Google
139.6
From 325 benchmarks. Plenty of data behind it.
121 ±8
Coding IQ 121.0, 95% interval 113–129. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.375cache read $0.0375$1.8751M
55.0
Good value (55.0). Cheaper per task than its benchmark results predict. Plenty of data behind it.
64 ±10
Chatty (64 ±10). Generates more tokens than models with similar coding results. Measured at high effort. A fair amount of data behind it.
158.8
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Gemini 3.8 Flash
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
global/priority
Inference by Google
139.6
From 325 benchmarks. Plenty of data behind it.
121 ±8
Coding IQ 121.0, 95% interval 113–129. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$1.35cache read $0.135$6.751M
55.0
Good value (55.0). Cheaper per task than its benchmark results predict. Plenty of data behind it.
64 ±10
Chatty (64 ±10). Generates more tokens than models with similar coding results. Measured at high effort. A fair amount of data behind it.
158.8
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Gemini 3.8 Flash
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
priority
Inference by Google
139.6
From 325 benchmarks. Plenty of data behind it.
121 ±8
Coding IQ 121.0, 95% interval 113–129. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$1.35cache read $0.135$6.751M
55.0
Good value (55.0). Cheaper per task than its benchmark results predict. Plenty of data behind it.
64 ±10
Chatty (64 ±10). Generates more tokens than models with similar coding results. Measured at high effort. A fair amount of data behind it.
158.8
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GPT-5.4 Pro
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Newer version
Newer versionGPT-5.5 ProWe compared the version names in our catalogue.
139.6
From 28 benchmarks. Plenty of data behind it.
Not enough data yet$30.00$180.001.1M
64.9
Good value (64.9). Cheaper per task than its benchmark results predict. A fair amount of data behind it.
Not measured
160.3
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GPT-5.4 Pro
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Newer version
Newer versionGPT-5.5 ProWe compared the version names in our catalogue.
openai/gpt-5.4-pro:batch · batch
139.6
From 28 benchmarks. Plenty of data behind it.
Not enough data yet$15.00$90.001.1M
64.9
Good value (64.9). Cheaper per task than its benchmark results predict. A fair amount of data behind it.
Not measured
160.3
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GPT-5.4 Pro
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Newer version
Newer versionGPT-5.5 ProWe compared the version names in our catalogue.
Inference by Microsoft Azure
139.6
From 28 benchmarks. Plenty of data behind it.
Not enough data yet$30.00$180.00922K
64.9
Good value (64.9). Cheaper per task than its benchmark results predict. A fair amount of data behind it.
Not measured
160.3
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GPT-5.4 Pro
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Newer version
Newer versionGPT-5.5 ProWe compared the version names in our catalogue.
Inference by OpenAI
139.6
From 28 benchmarks. Plenty of data behind it.
Not enough data yet$30.00$180.00922K
64.9
Good value (64.9). Cheaper per task than its benchmark results predict. A fair amount of data behind it.
Not measured
160.3
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Kimi K3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
Newer version
Newer versionKimi K3.1We compared the version names in our catalogue.
139.2
From 417 benchmarks. Plenty of data behind it.
124 ±7
Coding IQ 123.8, 95% interval 117–131. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.192cache read $0.182$16.001M
60.2
Good value (60.2). Cheaper per task than its benchmark results predict. Plenty of data behind it.
55 ±9
Typical (55 ±9). Generates about as many tokens as models with similar coding results. Measured at max effort. Little data behind it, so treat it as a rough estimate.
158.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Kimi K3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
Newer version
Newer versionKimi K3.1We compared the version names in our catalogue.
moonshotai/kimi-k3:batch · batch
139.2
From 417 benchmarks. Plenty of data behind it.
124 ±7
Coding IQ 123.8, 95% interval 117–131. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$2.28cache read $0.228$11.401M
60.2
Good value (60.2). Cheaper per task than its benchmark results predict. Plenty of data behind it.
55 ±9
Typical (55 ±9). Generates about as many tokens as models with similar coding results. Measured at max effort. Little data behind it, so treat it as a rough estimate.
158.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Kimi K3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
Newer version
Newer versionKimi K3.1We compared the version names in our catalogue.
Inference by Alibaba Cloud
139.2
From 417 benchmarks. Plenty of data behind it.
124 ±7
Coding IQ 123.8, 95% interval 117–131. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$3.00cache read $0.30$15.001M
60.2
Good value (60.2). Cheaper per task than its benchmark results predict. Plenty of data behind it.
55 ±9
Typical (55 ±9). Generates about as many tokens as models with similar coding results. Measured at max effort. Little data behind it, so treat it as a rough estimate.
158.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Kimi K3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
Newer version
Newer versionKimi K3.1We compared the version names in our catalogue.
us-east-1
Inference by Amazon Web Services
139.2
From 417 benchmarks. Plenty of data behind it.
124 ±7
Coding IQ 123.8, 95% interval 117–131. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$3.30cache read $0.33$16.50cache write $4.1251M
60.2
Good value (60.2). Cheaper per task than its benchmark results predict. Plenty of data behind it.
55 ±9
Typical (55 ±9). Generates about as many tokens as models with similar coding results. Measured at max effort. Little data behind it, so treat it as a rough estimate.
158.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Kimi K3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
Newer version
Newer versionKimi K3.1We compared the version names in our catalogue.
us-east-2
Inference by Amazon Web Services
139.2
From 417 benchmarks. Plenty of data behind it.
124 ±7
Coding IQ 123.8, 95% interval 117–131. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$3.00cache read $0.30$15.00cache write $3.751M
60.2
Good value (60.2). Cheaper per task than its benchmark results predict. Plenty of data behind it.
55 ±9
Typical (55 ±9). Generates about as many tokens as models with similar coding results. Measured at max effort. Little data behind it, so treat it as a rough estimate.
158.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Kimi K3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
Newer version
Newer versionKimi K3.1We compared the version names in our catalogue.
fp8
Inference by Baseten
139.2
From 417 benchmarks. Plenty of data behind it.
124 ±7
Coding IQ 123.8, 95% interval 117–131. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$3.00cache read $0.30$15.001M
60.2
Good value (60.2). Cheaper per task than its benchmark results predict. Plenty of data behind it.
55 ±9
Typical (55 ±9). Generates about as many tokens as models with similar coding results. Measured at max effort. Little data behind it, so treat it as a rough estimate.
158.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Kimi K3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
Newer version
Newer versionKimi K3.1We compared the version names in our catalogue.
mxfp4
Inference by Chutes
139.2
From 417 benchmarks. Plenty of data behind it.
124 ±7
Coding IQ 123.8, 95% interval 117–131. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$1.85cache read $0.185$9.501M
60.2
Good value (60.2). Cheaper per task than its benchmark results predict. Plenty of data behind it.
55 ±9
Typical (55 ±9). Generates about as many tokens as models with similar coding results. Measured at max effort. Little data behind it, so treat it as a rough estimate.
158.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Kimi K3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
Newer version
Newer versionKimi K3.1We compared the version names in our catalogue.
moonshotai/kimi-k3:batch · batch+bf16 · bf16
Inference by DeepInfra
139.2
From 417 benchmarks. Plenty of data behind it.
124 ±7
Coding IQ 123.8, 95% interval 117–131. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$2.28cache read $0.228$11.401M
60.2
Good value (60.2). Cheaper per task than its benchmark results predict. Plenty of data behind it.
55 ±9
Typical (55 ±9). Generates about as many tokens as models with similar coding results. Measured at max effort. Little data behind it, so treat it as a rough estimate.
158.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Kimi K3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
Newer version
Newer versionKimi K3.1We compared the version names in our catalogue.
mxfp4
Inference by DeepInfra
139.2
From 417 benchmarks. Plenty of data behind it.
124 ±7
Coding IQ 123.8, 95% interval 117–131. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$2.85cache read $0.285$14.251M
60.2
Good value (60.2). Cheaper per task than its benchmark results predict. Plenty of data behind it.
55 ±9
Typical (55 ±9). Generates about as many tokens as models with similar coding results. Measured at max effort. Little data behind it, so treat it as a rough estimate.
158.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Kimi K3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
Newer version
Newer versionKimi K3.1We compared the version names in our catalogue.
Inference by Digitalocean
139.2
From 417 benchmarks. Plenty of data behind it.
124 ±7
Coding IQ 123.8, 95% interval 117–131. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$2.55cache read $0.255$12.951M
60.2
Good value (60.2). Cheaper per task than its benchmark results predict. Plenty of data behind it.
55 ±9
Typical (55 ±9). Generates about as many tokens as models with similar coding results. Measured at max effort. Little data behind it, so treat it as a rough estimate.
158.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Kimi K3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
Newer version
Newer versionKimi K3.1We compared the version names in our catalogue.
Inference by Fireworks AI
139.2
From 417 benchmarks. Plenty of data behind it.
124 ±7
Coding IQ 123.8, 95% interval 117–131. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$3.00cache read $0.30$15.001M
60.2
Good value (60.2). Cheaper per task than its benchmark results predict. Plenty of data behind it.
55 ±9
Typical (55 ±9). Generates about as many tokens as models with similar coding results. Measured at max effort. Little data behind it, so treat it as a rough estimate.
158.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Kimi K3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
Newer version
Newer versionKimi K3.1We compared the version names in our catalogue.
fast
Inference by Fireworks AI
139.2
From 417 benchmarks. Plenty of data behind it.
124 ±7
Coding IQ 123.8, 95% interval 117–131. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$4.50cache read $0.45$22.501M
60.2
Good value (60.2). Cheaper per task than its benchmark results predict. Plenty of data behind it.
55 ±9
Typical (55 ±9). Generates about as many tokens as models with similar coding results. Measured at max effort. Little data behind it, so treat it as a rough estimate.
158.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Kimi K3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
Newer version
Newer versionKimi K3.1We compared the version names in our catalogue.
us
Inference by Fireworks AI
139.2
From 417 benchmarks. Plenty of data behind it.
124 ±7
Coding IQ 123.8, 95% interval 117–131. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$4.50cache read $0.45$22.501M
60.2
Good value (60.2). Cheaper per task than its benchmark results predict. Plenty of data behind it.
55 ±9
Typical (55 ±9). Generates about as many tokens as models with similar coding results. Measured at max effort. Little data behind it, so treat it as a rough estimate.
158.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Kimi K3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
Newer version
Newer versionKimi K3.1We compared the version names in our catalogue.
fast · fp4
Inference by Inferencenet
139.2
From 417 benchmarks. Plenty of data behind it.
124 ±7
Coding IQ 123.8, 95% interval 117–131. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.45cache read $0.363$18.001M
60.2
Good value (60.2). Cheaper per task than its benchmark results predict. Plenty of data behind it.
55 ±9
Typical (55 ±9). Generates about as many tokens as models with similar coding results. Measured at max effort. Little data behind it, so treat it as a rough estimate.
158.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Kimi K3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
Newer version
Newer versionKimi K3.1We compared the version names in our catalogue.
fp4
Inference by Inferencenet
139.2
From 417 benchmarks. Plenty of data behind it.
124 ±7
Coding IQ 123.8, 95% interval 117–131. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.245cache read $0.214$15.751M
60.2
Good value (60.2). Cheaper per task than its benchmark results predict. Plenty of data behind it.
55 ±9
Typical (55 ±9). Generates about as many tokens as models with similar coding results. Measured at max effort. Little data behind it, so treat it as a rough estimate.
158.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Kimi K3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
Newer version
Newer versionKimi K3.1We compared the version names in our catalogue.
us · fp4
Inference by Inferencenet
139.2
From 417 benchmarks. Plenty of data behind it.
124 ±7
Coding IQ 123.8, 95% interval 117–131. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.245cache read $0.214$15.751M
60.2
Good value (60.2). Cheaper per task than its benchmark results predict. Plenty of data behind it.
55 ±9
Typical (55 ±9). Generates about as many tokens as models with similar coding results. Measured at max effort. Little data behind it, so treat it as a rough estimate.
158.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Kimi K3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
Newer version
Newer versionKimi K3.1We compared the version names in our catalogue.
us/fast · fp4
Inference by Inferencenet
139.2
From 417 benchmarks. Plenty of data behind it.
124 ±7
Coding IQ 123.8, 95% interval 117–131. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.45cache read $0.363$18.001M
60.2
Good value (60.2). Cheaper per task than its benchmark results predict. Plenty of data behind it.
55 ±9
Typical (55 ±9). Generates about as many tokens as models with similar coding results. Measured at max effort. Little data behind it, so treat it as a rough estimate.
158.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Kimi K3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
Newer version
Newer versionKimi K3.1We compared the version names in our catalogue.
fp4
Inference by Ionet
139.2
From 417 benchmarks. Plenty of data behind it.
124 ±7
Coding IQ 123.8, 95% interval 117–131. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.60cache read $0.12$11.501M
60.2
Good value (60.2). Cheaper per task than its benchmark results predict. Plenty of data behind it.
55 ±9
Typical (55 ±9). Generates about as many tokens as models with similar coding results. Measured at max effort. Little data behind it, so treat it as a rough estimate.
158.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Kimi K3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
Newer version
Newer versionKimi K3.1We compared the version names in our catalogue.
Inference by Kt
139.2
From 417 benchmarks. Plenty of data behind it.
124 ±7
Coding IQ 123.8, 95% interval 117–131. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.45cache read $0.30$15.001M
60.2
Good value (60.2). Cheaper per task than its benchmark results predict. Plenty of data behind it.
55 ±9
Typical (55 ±9). Generates about as many tokens as models with similar coding results. Measured at max effort. Little data behind it, so treat it as a rough estimate.
158.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Kimi K3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
Newer version
Newer versionKimi K3.1We compared the version names in our catalogue.
fp4
Inference by Makora
139.2
From 417 benchmarks. Plenty of data behind it.
124 ±7
Coding IQ 123.8, 95% interval 117–131. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$1.00cache read $0.50$18.001M
60.2
Good value (60.2). Cheaper per task than its benchmark results predict. Plenty of data behind it.
55 ±9
Typical (55 ±9). Generates about as many tokens as models with similar coding results. Measured at max effort. Little data behind it, so treat it as a rough estimate.
158.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Kimi K3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
Newer version
Newer versionKimi K3.1We compared the version names in our catalogue.
mxfp4
Inference by Modal
139.2
From 417 benchmarks. Plenty of data behind it.
124 ±7
Coding IQ 123.8, 95% interval 117–131. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$3.00cache read $0.30$15.001M
60.2
Good value (60.2). Cheaper per task than its benchmark results predict. Plenty of data behind it.
55 ±9
Typical (55 ±9). Generates about as many tokens as models with similar coding results. Measured at max effort. Little data behind it, so treat it as a rough estimate.
158.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Kimi K3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
Newer version
Newer versionKimi K3.1We compared the version names in our catalogue.
Inference by Moonshot AI
139.2
From 417 benchmarks. Plenty of data behind it.
124 ±7
Coding IQ 123.8, 95% interval 117–131. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$3.00cache read $0.30$15.001M
60.2
Good value (60.2). Cheaper per task than its benchmark results predict. Plenty of data behind it.
55 ±9
Typical (55 ±9). Generates about as many tokens as models with similar coding results. Measured at max effort. Little data behind it, so treat it as a rough estimate.
158.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Kimi K3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
Newer version
Newer versionKimi K3.1We compared the version names in our catalogue.
fp4
Inference by Nebius AI Studio
139.2
From 417 benchmarks. Plenty of data behind it.
124 ±7
Coding IQ 123.8, 95% interval 117–131. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$3.00cache read $0.30$15.001M
60.2
Good value (60.2). Cheaper per task than its benchmark results predict. Plenty of data behind it.
55 ±9
Typical (55 ±9). Generates about as many tokens as models with similar coding results. Measured at max effort. Little data behind it, so treat it as a rough estimate.
158.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Kimi K3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
Newer version
Newer versionKimi K3.1We compared the version names in our catalogue.
fast · fp4
Inference by Parasail
139.2
From 417 benchmarks. Plenty of data behind it.
124 ±7
Coding IQ 123.8, 95% interval 117–131. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$4.50cache read $0.45$22.501M
60.2
Good value (60.2). Cheaper per task than its benchmark results predict. Plenty of data behind it.
55 ±9
Typical (55 ±9). Generates about as many tokens as models with similar coding results. Measured at max effort. Little data behind it, so treat it as a rough estimate.
158.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Kimi K3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
Newer version
Newer versionKimi K3.1We compared the version names in our catalogue.
fp4
Inference by Parasail
139.2
From 417 benchmarks. Plenty of data behind it.
124 ±7
Coding IQ 123.8, 95% interval 117–131. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$3.00cache read $0.30$15.001M
60.2
Good value (60.2). Cheaper per task than its benchmark results predict. Plenty of data behind it.
55 ±9
Typical (55 ±9). Generates about as many tokens as models with similar coding results. Measured at max effort. Little data behind it, so treat it as a rough estimate.
158.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Kimi K3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
Newer version
Newer versionKimi K3.1We compared the version names in our catalogue.
fp4
Inference by Relace
139.2
From 417 benchmarks. Plenty of data behind it.
124 ±7
Coding IQ 123.8, 95% interval 117–131. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.1824cache read $0.18$11.0061M
60.2
Good value (60.2). Cheaper per task than its benchmark results predict. Plenty of data behind it.
55 ±9
Typical (55 ±9). Generates about as many tokens as models with similar coding results. Measured at max effort. Little data behind it, so treat it as a rough estimate.
158.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Kimi K3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
Newer version
Newer versionKimi K3.1We compared the version names in our catalogue.
Inference by River
139.2
From 417 benchmarks. Plenty of data behind it.
124 ±7
Coding IQ 123.8, 95% interval 117–131. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$3.00cache read $0.30$15.001M
60.2
Good value (60.2). Cheaper per task than its benchmark results predict. Plenty of data behind it.
55 ±9
Typical (55 ±9). Generates about as many tokens as models with similar coding results. Measured at max effort. Little data behind it, so treat it as a rough estimate.
158.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Kimi K3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
Newer version
Newer versionKimi K3.1We compared the version names in our catalogue.
fp4
Inference by Sail Research
139.2
From 417 benchmarks. Plenty of data behind it.
124 ±7
Coding IQ 123.8, 95% interval 117–131. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.345cache read $0.1725$16.581M
60.2
Good value (60.2). Cheaper per task than its benchmark results predict. Plenty of data behind it.
55 ±9
Typical (55 ±9). Generates about as many tokens as models with similar coding results. Measured at max effort. Little data behind it, so treat it as a rough estimate.
158.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Kimi K3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
Newer version
Newer versionKimi K3.1We compared the version names in our catalogue.
Inference by Together AI
139.2
From 417 benchmarks. Plenty of data behind it.
124 ±7
Coding IQ 123.8, 95% interval 117–131. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$3.00cache read $0.30$15.001M
60.2
Good value (60.2). Cheaper per task than its benchmark results predict. Plenty of data behind it.
55 ±9
Typical (55 ±9). Generates about as many tokens as models with similar coding results. Measured at max effort. Little data behind it, so treat it as a rough estimate.
158.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Kimi K3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
Newer version
Newer versionKimi K3.1We compared the version names in our catalogue.
mxfp4
Inference by Wafer
139.2
From 417 benchmarks. Plenty of data behind it.
124 ±7
Coding IQ 123.8, 95% interval 117–131. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.192cache read $0.182$16.001M
60.2
Good value (60.2). Cheaper per task than its benchmark results predict. Plenty of data behind it.
55 ±9
Typical (55 ±9). Generates about as many tokens as models with similar coding results. Measured at max effort. Little data behind it, so treat it as a rough estimate.
158.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Kimi K3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
Newer version
Newer versionKimi K3.1We compared the version names in our catalogue.
fast · mxfp4
Inference by Wafer
139.2
From 417 benchmarks. Plenty of data behind it.
124 ±7
Coding IQ 123.8, 95% interval 117–131. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.153cache read $0.152$16.001M
60.2
Good value (60.2). Cheaper per task than its benchmark results predict. Plenty of data behind it.
55 ±9
Typical (55 ±9). Generates about as many tokens as models with similar coding results. Measured at max effort. Little data behind it, so treat it as a rough estimate.
158.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Kimi K3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
Newer version
Newer versionKimi K3.1We compared the version names in our catalogue.
us · mxfp4
Inference by Wafer
139.2
From 417 benchmarks. Plenty of data behind it.
124 ±7
Coding IQ 123.8, 95% interval 117–131. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.276cache read $0.13$9.601M
60.2
Good value (60.2). Cheaper per task than its benchmark results predict. Plenty of data behind it.
55 ±9
Typical (55 ±9). Generates about as many tokens as models with similar coding results. Measured at max effort. Little data behind it, so treat it as a rough estimate.
158.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Gemini 3.7 Flash
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Newer version
Newer versionGemini 3.8 FlashWe compared release dates and version names in our catalogue.
138.8
From 267 benchmarks. Plenty of data behind it.
121 ±9
Coding IQ 121.3, 95% interval 113–130. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.75cache read $0.075$3.751M
60.3
Good value (60.3). Cheaper per task than its benchmark results predict. Plenty of data behind it.
59 ±12
Chatty (59 ±12). Generates more tokens than models with similar coding results. Measured at high effort. A fair amount of data behind it.
158.4
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Gemini 3.7 Flash
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Newer version
Newer versionGemini 3.8 FlashWe compared release dates and version names in our catalogue.
google/gemini-3.7-flash:batch · batch
138.8
From 267 benchmarks. Plenty of data behind it.
121 ±9
Coding IQ 121.3, 95% interval 113–130. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.375cache read $0.0375$1.8751M
60.3
Good value (60.3). Cheaper per task than its benchmark results predict. Plenty of data behind it.
59 ±12
Chatty (59 ±12). Generates more tokens than models with similar coding results. Measured at high effort. A fair amount of data behind it.
158.4
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Gemini 3.7 Flash
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Newer version
Newer versionGemini 3.8 FlashWe compared release dates and version names in our catalogue.
Inference by Google
138.8
From 267 benchmarks. Plenty of data behind it.
121 ±9
Coding IQ 121.3, 95% interval 113–130. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.75cache read $0.075$3.751M
60.3
Good value (60.3). Cheaper per task than its benchmark results predict. Plenty of data behind it.
59 ±12
Chatty (59 ±12). Generates more tokens than models with similar coding results. Measured at high effort. A fair amount of data behind it.
158.4
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Gemini 3.7 Flash
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Newer version
Newer versionGemini 3.8 FlashWe compared release dates and version names in our catalogue.
google/gemini-3.7-flash:batch · batch+global
Inference by Google
138.8
From 267 benchmarks. Plenty of data behind it.
121 ±9
Coding IQ 121.3, 95% interval 113–130. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.375cache read $0.0375$1.8751M
60.3
Good value (60.3). Cheaper per task than its benchmark results predict. Plenty of data behind it.
59 ±12
Chatty (59 ±12). Generates more tokens than models with similar coding results. Measured at high effort. A fair amount of data behind it.
158.4
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Gemini 3.7 Flash
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Newer version
Newer versionGemini 3.8 FlashWe compared release dates and version names in our catalogue.
flex
Inference by Google
138.8
From 267 benchmarks. Plenty of data behind it.
121 ±9
Coding IQ 121.3, 95% interval 113–130. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.375cache read $0.0375$1.8751M
60.3
Good value (60.3). Cheaper per task than its benchmark results predict. Plenty of data behind it.
59 ±12
Chatty (59 ±12). Generates more tokens than models with similar coding results. Measured at high effort. A fair amount of data behind it.
158.4
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Gemini 3.7 Flash
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Newer version
Newer versionGemini 3.8 FlashWe compared release dates and version names in our catalogue.
global
Inference by Google
138.8
From 267 benchmarks. Plenty of data behind it.
121 ±9
Coding IQ 121.3, 95% interval 113–130. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.75cache read $0.075$3.751M
60.3
Good value (60.3). Cheaper per task than its benchmark results predict. Plenty of data behind it.
59 ±12
Chatty (59 ±12). Generates more tokens than models with similar coding results. Measured at high effort. A fair amount of data behind it.
158.4
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Gemini 3.7 Flash
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Newer version
Newer versionGemini 3.8 FlashWe compared release dates and version names in our catalogue.
global/flex
Inference by Google
138.8
From 267 benchmarks. Plenty of data behind it.
121 ±9
Coding IQ 121.3, 95% interval 113–130. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.375cache read $0.0375$1.8751M
60.3
Good value (60.3). Cheaper per task than its benchmark results predict. Plenty of data behind it.
59 ±12
Chatty (59 ±12). Generates more tokens than models with similar coding results. Measured at high effort. A fair amount of data behind it.
158.4
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Gemini 3.7 Flash
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Newer version
Newer versionGemini 3.8 FlashWe compared release dates and version names in our catalogue.
global/priority
Inference by Google
138.8
From 267 benchmarks. Plenty of data behind it.
121 ±9
Coding IQ 121.3, 95% interval 113–130. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$1.35cache read $0.135$6.751M
60.3
Good value (60.3). Cheaper per task than its benchmark results predict. Plenty of data behind it.
59 ±12
Chatty (59 ±12). Generates more tokens than models with similar coding results. Measured at high effort. A fair amount of data behind it.
158.4
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Gemini 3.7 Flash
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Newer version
Newer versionGemini 3.8 FlashWe compared release dates and version names in our catalogue.
priority
Inference by Google
138.8
From 267 benchmarks. Plenty of data behind it.
121 ±9
Coding IQ 121.3, 95% interval 113–130. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$1.35cache read $0.135$6.751M
60.3
Good value (60.3). Cheaper per task than its benchmark results predict. Plenty of data behind it.
59 ±12
Chatty (59 ±12). Generates more tokens than models with similar coding results. Measured at high effort. A fair amount of data behind it.
158.4
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Qwen3.8 Max
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Newer version
Newer versionQwen4 MaxWe compared release dates and version names in our catalogue.
qwen/qwen3.8-2.4t-a95b
138.7
From 312 benchmarks. Plenty of data behind it.
117 ±5
Coding IQ 117.2, 95% interval 112–122. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$2.00cache read $0.25$6.00262K
49.2
Fair value (49.2). Costs per task about what its benchmark results predict. Plenty of data behind it.
62 ±9
Chatty (62 ±9). Generates more tokens than models with similar coding results. Measured at max effort. Little data behind it, so treat it as a rough estimate.
157.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Qwen3.8 Max
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Newer version
Newer versionQwen4 MaxWe compared release dates and version names in our catalogue.
qwen/qwen3.8-max-0902 · 0902
138.7
From 312 benchmarks. Plenty of data behind it.
117 ±5
Coding IQ 117.2, 95% interval 112–122. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$2.00cache read $0.25$6.001M
49.2
Fair value (49.2). Costs per task about what its benchmark results predict. Plenty of data behind it.
62 ±9
Chatty (62 ±9). Generates more tokens than models with similar coding results. Measured at max effort. Little data behind it, so treat it as a rough estimate.
157.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Qwen3.8 Max
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Newer version
Newer versionQwen4 MaxWe compared release dates and version names in our catalogue.
qwen/qwen3.8-max-0902:batch · batch+0902
138.7
From 312 benchmarks. Plenty of data behind it.
117 ±5
Coding IQ 117.2, 95% interval 112–122. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$1.15$3.45262K
49.2
Fair value (49.2). Costs per task about what its benchmark results predict. Plenty of data behind it.
62 ±9
Chatty (62 ±9). Generates more tokens than models with similar coding results. Measured at max effort. Little data behind it, so treat it as a rough estimate.
157.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Qwen3.8 Max
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Newer version
Newer versionQwen4 MaxWe compared release dates and version names in our catalogue.
qwen/qwen3.8-2.4t-a95b
Inference by Alibaba Cloud
138.7
From 312 benchmarks. Plenty of data behind it.
117 ±5
Coding IQ 117.2, 95% interval 112–122. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$2.00cache read $0.25$6.00cache write $2.501M
49.2
Fair value (49.2). Costs per task about what its benchmark results predict. Plenty of data behind it.
62 ±9
Chatty (62 ±9). Generates more tokens than models with similar coding results. Measured at max effort. Little data behind it, so treat it as a rough estimate.
157.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Qwen3.8 Max
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Newer version
Newer versionQwen4 MaxWe compared release dates and version names in our catalogue.
qwen/qwen3.8-max-0902 · 0902
Inference by Alibaba Cloud
138.7
From 312 benchmarks. Plenty of data behind it.
117 ±5
Coding IQ 117.2, 95% interval 112–122. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$2.00cache read $0.25$6.00cache write $2.50984K
49.2
Fair value (49.2). Costs per task about what its benchmark results predict. Plenty of data behind it.
62 ±9
Chatty (62 ±9). Generates more tokens than models with similar coding results. Measured at max effort. Little data behind it, so treat it as a rough estimate.
157.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Qwen3.8 Max
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Newer version
Newer versionQwen4 MaxWe compared release dates and version names in our catalogue.
qwen/qwen3.8-max-0902:batch · batch+0902
Inference by Alibaba Cloud
138.7
From 312 benchmarks. Plenty of data behind it.
117 ±5
Coding IQ 117.2, 95% interval 112–122. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$1.15$3.45246K
49.2
Fair value (49.2). Costs per task about what its benchmark results predict. Plenty of data behind it.
62 ±9
Chatty (62 ±9). Generates more tokens than models with similar coding results. Measured at max effort. Little data behind it, so treat it as a rough estimate.
157.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Qwen3.8 Max
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Newer version
Newer versionQwen4 MaxWe compared release dates and version names in our catalogue.
qwen/qwen3.8-max-0902 · us-east-1+0902
Inference by Alibaba Cloud
138.7
From 312 benchmarks. Plenty of data behind it.
117 ±5
Coding IQ 117.2, 95% interval 112–122. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$2.00cache read $0.25$6.00984K
49.2
Fair value (49.2). Costs per task about what its benchmark results predict. Plenty of data behind it.
62 ±9
Chatty (62 ±9). Generates more tokens than models with similar coding results. Measured at max effort. Little data behind it, so treat it as a rough estimate.
157.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Qwen3.8 Max
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Newer version
Newer versionQwen4 MaxWe compared release dates and version names in our catalogue.
qwen/qwen3.8-2.4t-a95b
Inference by Fireworks AI
138.7
From 312 benchmarks. Plenty of data behind it.
117 ±5
Coding IQ 117.2, 95% interval 112–122. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$2.00cache read $0.25$6.001M
49.2
Fair value (49.2). Costs per task about what its benchmark results predict. Plenty of data behind it.
62 ±9
Chatty (62 ±9). Generates more tokens than models with similar coding results. Measured at max effort. Little data behind it, so treat it as a rough estimate.
157.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Qwen3.8 Max
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Newer version
Newer versionQwen4 MaxWe compared release dates and version names in our catalogue.
qwen/qwen3.8-2.4t-a95b
Inference by Novita AI
138.7
From 312 benchmarks. Plenty of data behind it.
117 ±5
Coding IQ 117.2, 95% interval 112–122. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$2.00cache read $0.25$6.001M
49.2
Fair value (49.2). Costs per task about what its benchmark results predict. Plenty of data behind it.
62 ±9
Chatty (62 ±9). Generates more tokens than models with similar coding results. Measured at max effort. Little data behind it, so treat it as a rough estimate.
157.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Qwen3.8 Max
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Newer version
Newer versionQwen4 MaxWe compared release dates and version names in our catalogue.
qwen/qwen3.8-2.4t-a95b · fp8
Inference by SiliconFlow
138.7
From 312 benchmarks. Plenty of data behind it.
117 ±5
Coding IQ 117.2, 95% interval 112–122. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$2.00cache read $0.25$6.001M
49.2
Fair value (49.2). Costs per task about what its benchmark results predict. Plenty of data behind it.
62 ±9
Chatty (62 ±9). Generates more tokens than models with similar coding results. Measured at max effort. Little data behind it, so treat it as a rough estimate.
157.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Qwen3.8 Max
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Newer version
Newer versionQwen4 MaxWe compared release dates and version names in our catalogue.
qwen/qwen3.8-2.4t-a95b
Inference by Venice
138.7
From 312 benchmarks. Plenty of data behind it.
117 ±5
Coding IQ 117.2, 95% interval 112–122. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$2.00cache read $0.25$6.00262K
49.2
Fair value (49.2). Costs per task about what its benchmark results predict. Plenty of data behind it.
62 ±9
Chatty (62 ±9). Generates more tokens than models with similar coding results. Measured at max effort. Little data behind it, so treat it as a rough estimate.
157.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Claude Opus 4.7
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Newer version
Newer versionClaude Opus 5.5We compared release dates and version names in our catalogue.
anthropic/claude-opus-4.7-fast · fast
138.2
From 354 benchmarks. Plenty of data behind it.
118 ±15
Coding IQ 118.2, 95% interval 103–134. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$30.00cache read $3.00$150.00cache write $37.50Not published by the provider
49.6
Fair value (49.6). Costs per task about what its benchmark results predict. Plenty of data behind it.
38 ±12
Terse (38 ±12). Generates fewer tokens than models with similar coding results. Measured at default effort. A fair amount of data behind it.
157.5
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Claude Opus 4.7
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Newer version
Newer versionClaude Opus 5.5We compared release dates and version names in our catalogue.
138.2
From 354 benchmarks. Plenty of data behind it.
118 ±15
Coding IQ 118.2, 95% interval 103–134. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$5.00cache read $0.50$25.00cache write $6.251M
49.6
Fair value (49.6). Costs per task about what its benchmark results predict. Plenty of data behind it.
38 ±12
Terse (38 ±12). Generates fewer tokens than models with similar coding results. Measured at default effort. A fair amount of data behind it.
157.5
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Claude Opus 4.7
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Newer version
Newer versionClaude Opus 5.5We compared release dates and version names in our catalogue.
anthropic/claude-opus-4.7:batch · batch
138.2
From 354 benchmarks. Plenty of data behind it.
118 ±15
Coding IQ 118.2, 95% interval 103–134. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$2.50cache read $0.25$12.50cache write $3.1251M
49.6
Fair value (49.6). Costs per task about what its benchmark results predict. Plenty of data behind it.
38 ±12
Terse (38 ±12). Generates fewer tokens than models with similar coding results. Measured at default effort. A fair amount of data behind it.
157.5
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Claude Opus 4.7
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Newer version
Newer versionClaude Opus 5.5We compared release dates and version names in our catalogue.
global
Inference by Amazon Web Services
138.2
From 354 benchmarks. Plenty of data behind it.
118 ±15
Coding IQ 118.2, 95% interval 103–134. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$5.00cache read $0.50$25.00cache write $6.251M
49.6
Fair value (49.6). Costs per task about what its benchmark results predict. Plenty of data behind it.
38 ±12
Terse (38 ±12). Generates fewer tokens than models with similar coding results. Measured at default effort. A fair amount of data behind it.
157.5
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Claude Opus 4.7
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Newer version
Newer versionClaude Opus 5.5We compared release dates and version names in our catalogue.
Inference by Anthropic
138.2
From 354 benchmarks. Plenty of data behind it.
118 ±15
Coding IQ 118.2, 95% interval 103–134. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$5.00cache read $0.50$25.00cache write $6.251M
49.6
Fair value (49.6). Costs per task about what its benchmark results predict. Plenty of data behind it.
38 ±12
Terse (38 ±12). Generates fewer tokens than models with similar coding results. Measured at default effort. A fair amount of data behind it.
157.5
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Claude Opus 4.7
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Newer version
Newer versionClaude Opus 5.5We compared release dates and version names in our catalogue.
global
Inference by Microsoft Azure
138.2
From 354 benchmarks. Plenty of data behind it.
118 ±15
Coding IQ 118.2, 95% interval 103–134. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$5.00cache read $0.50$25.00cache write $6.251M
49.6
Fair value (49.6). Costs per task about what its benchmark results predict. Plenty of data behind it.
38 ±12
Terse (38 ±12). Generates fewer tokens than models with similar coding results. Measured at default effort. A fair amount of data behind it.
157.5
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Claude Opus 4.7
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Newer version
Newer versionClaude Opus 5.5We compared release dates and version names in our catalogue.
Inference by Claude Platform On AWS
138.2
From 354 benchmarks. Plenty of data behind it.
118 ±15
Coding IQ 118.2, 95% interval 103–134. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$5.00cache read $0.50$25.00cache write $6.251M
49.6
Fair value (49.6). Costs per task about what its benchmark results predict. Plenty of data behind it.
38 ±12
Terse (38 ±12). Generates fewer tokens than models with similar coding results. Measured at default effort. A fair amount of data behind it.
157.5
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Claude Opus 4.7
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Newer version
Newer versionClaude Opus 5.5We compared release dates and version names in our catalogue.
global
Inference by Google
138.2
From 354 benchmarks. Plenty of data behind it.
118 ±15
Coding IQ 118.2, 95% interval 103–134. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$5.00cache read $0.50$25.00cache write $6.251M
49.6
Fair value (49.6). Costs per task about what its benchmark results predict. Plenty of data behind it.
38 ±12
Terse (38 ±12). Generates fewer tokens than models with similar coding results. Measured at default effort. A fair amount of data behind it.
157.5
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GLM 5.3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
138.2
From 318 benchmarks. Plenty of data behind it.
121 ±9
Coding IQ 121.0, 95% interval 112–130. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.0699cache read $0.0689$1.401M
57.7
Good value (57.7). Cheaper per task than its benchmark results predict. Plenty of data behind it.
65 ±10
Chatty (65 ±10). Generates more tokens than models with similar coding results. Measured at max effort. A fair amount of data behind it.
157.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GLM 5.3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
z-ai/glm-5.3:batch · batch
138.2
From 318 benchmarks. Plenty of data behind it.
121 ±9
Coding IQ 121.0, 95% interval 112–130. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.28cache read $0.061$1.2491M
57.7
Good value (57.7). Cheaper per task than its benchmark results predict. Plenty of data behind it.
65 ±10
Chatty (65 ±10). Generates more tokens than models with similar coding results. Measured at max effort. A fair amount of data behind it.
157.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GLM 5.3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
fp4
Inference by Akash
138.2
From 318 benchmarks. Plenty of data behind it.
121 ±9
Coding IQ 121.0, 95% interval 112–130. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.19$4.401M
57.7
Good value (57.7). Cheaper per task than its benchmark results predict. Plenty of data behind it.
65 ±10
Chatty (65 ±10). Generates more tokens than models with similar coding results. Measured at max effort. A fair amount of data behind it.
157.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GLM 5.3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
Inference by Alibaba Cloud
138.2
From 318 benchmarks. Plenty of data behind it.
121 ±9
Coding IQ 121.0, 95% interval 112–130. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$1.19cache read $0.238$3.741M
57.7
Good value (57.7). Cheaper per task than its benchmark results predict. Plenty of data behind it.
65 ±10
Chatty (65 ±10). Generates more tokens than models with similar coding results. Measured at max effort. A fair amount of data behind it.
157.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GLM 5.3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
prime
Inference by Alibaba Cloud
138.2
From 318 benchmarks. Plenty of data behind it.
121 ±9
Coding IQ 121.0, 95% interval 112–130. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$1.68cache read $0.336$5.281M
57.7
Good value (57.7). Cheaper per task than its benchmark results predict. Plenty of data behind it.
65 ±10
Chatty (65 ±10). Generates more tokens than models with similar coding results. Measured at max effort. A fair amount of data behind it.
157.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GLM 5.3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
us-east-1
Inference by Amazon Web Services
138.2
From 318 benchmarks. Plenty of data behind it.
121 ±9
Coding IQ 121.0, 95% interval 112–130. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$1.68cache read $0.312$5.28cache write $2.101M
57.7
Good value (57.7). Cheaper per task than its benchmark results predict. Plenty of data behind it.
65 ±10
Chatty (65 ±10). Generates more tokens than models with similar coding results. Measured at max effort. A fair amount of data behind it.
157.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GLM 5.3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
fp8
Inference by Baidu
138.2
From 318 benchmarks. Plenty of data behind it.
121 ±9
Coding IQ 121.0, 95% interval 112–130. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$1.40cache read $0.26$4.401M
57.7
Good value (57.7). Cheaper per task than its benchmark results predict. Plenty of data behind it.
65 ±10
Chatty (65 ±10). Generates more tokens than models with similar coding results. Measured at max effort. A fair amount of data behind it.
157.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GLM 5.3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
fast · fp8
Inference by Baseten
138.2
From 318 benchmarks. Plenty of data behind it.
121 ±9
Coding IQ 121.0, 95% interval 112–130. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$2.10cache read $0.21$6.601M
57.7
Good value (57.7). Cheaper per task than its benchmark results predict. Plenty of data behind it.
65 ±10
Chatty (65 ±10). Generates more tokens than models with similar coding results. Measured at max effort. A fair amount of data behind it.
157.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GLM 5.3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
fp4
Inference by Baseten
138.2
From 318 benchmarks. Plenty of data behind it.
121 ±9
Coding IQ 121.0, 95% interval 112–130. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$1.40cache read $0.14$4.401M
57.7
Good value (57.7). Cheaper per task than its benchmark results predict. Plenty of data behind it.
65 ±10
Chatty (65 ±10). Generates more tokens than models with similar coding results. Measured at max effort. A fair amount of data behind it.
157.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GLM 5.3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
fp4/us · fp4
Inference by Baseten
138.2
From 318 benchmarks. Plenty of data behind it.
121 ±9
Coding IQ 121.0, 95% interval 112–130. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$1.40cache read $0.14$4.401M
57.7
Good value (57.7). Cheaper per task than its benchmark results predict. Plenty of data behind it.
65 ±10
Chatty (65 ±10). Generates more tokens than models with similar coding results. Measured at max effort. A fair amount of data behind it.
157.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GLM 5.3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
us/fast · fp8
Inference by Baseten
138.2
From 318 benchmarks. Plenty of data behind it.
121 ±9
Coding IQ 121.0, 95% interval 112–130. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$2.10cache read $0.21$6.601M
57.7
Good value (57.7). Cheaper per task than its benchmark results predict. Plenty of data behind it.
65 ±10
Chatty (65 ±10). Generates more tokens than models with similar coding results. Measured at max effort. A fair amount of data behind it.
157.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GLM 5.3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
Inference by Cloudflare Workers AI
138.2
From 318 benchmarks. Plenty of data behind it.
121 ±9
Coding IQ 121.0, 95% interval 112–130. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.70cache read $0.13$2.201M
57.7
Good value (57.7). Cheaper per task than its benchmark results predict. Plenty of data behind it.
65 ±10
Chatty (65 ±10). Generates more tokens than models with similar coding results. Measured at max effort. A fair amount of data behind it.
157.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GLM 5.3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
nvfp4
Inference by Cohere
138.2
From 318 benchmarks. Plenty of data behind it.
121 ±9
Coding IQ 121.0, 95% interval 112–130. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.50cache read $0.10$3.501M
57.7
Good value (57.7). Cheaper per task than its benchmark results predict. Plenty of data behind it.
65 ±10
Chatty (65 ±10). Generates more tokens than models with similar coding results. Measured at max effort. A fair amount of data behind it.
157.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GLM 5.3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
nvfp4
Inference by Coreweave
138.2
From 318 benchmarks. Plenty of data behind it.
121 ±9
Coding IQ 121.0, 95% interval 112–130. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.75cache read $0.19$3.001M
57.7
Good value (57.7). Cheaper per task than its benchmark results predict. Plenty of data behind it.
65 ±10
Chatty (65 ±10). Generates more tokens than models with similar coding results. Measured at max effort. A fair amount of data behind it.
157.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GLM 5.3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
fp4
Inference by Crusoe
138.2
From 318 benchmarks. Plenty of data behind it.
121 ±9
Coding IQ 121.0, 95% interval 112–130. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$1.40cache read $0.26$4.401M
57.7
Good value (57.7). Cheaper per task than its benchmark results predict. Plenty of data behind it.
65 ±10
Chatty (65 ±10). Generates more tokens than models with similar coding results. Measured at max effort. A fair amount of data behind it.
157.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GLM 5.3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
fp4
Inference by Decart
138.2
From 318 benchmarks. Plenty of data behind it.
121 ±9
Coding IQ 121.0, 95% interval 112–130. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.56cache read $0.10$1.811M
57.7
Good value (57.7). Cheaper per task than its benchmark results predict. Plenty of data behind it.
65 ±10
Chatty (65 ±10). Generates more tokens than models with similar coding results. Measured at max effort. A fair amount of data behind it.
157.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GLM 5.3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
z-ai/glm-5.3:batch · batch+fp4 · fp4
Inference by DeepInfra
138.2
From 318 benchmarks. Plenty of data behind it.
121 ±9
Coding IQ 121.0, 95% interval 112–130. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.45cache read $0.10$2.001M
57.7
Good value (57.7). Cheaper per task than its benchmark results predict. Plenty of data behind it.
65 ±10
Chatty (65 ±10). Generates more tokens than models with similar coding results. Measured at max effort. A fair amount of data behind it.
157.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GLM 5.3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
fp4
Inference by DeepInfra
138.2
From 318 benchmarks. Plenty of data behind it.
121 ±9
Coding IQ 121.0, 95% interval 112–130. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.5625cache read $0.125$2.501M
57.7
Good value (57.7). Cheaper per task than its benchmark results predict. Plenty of data behind it.
65 ±10
Chatty (65 ±10). Generates more tokens than models with similar coding results. Measured at max effort. A fair amount of data behind it.
157.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GLM 5.3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
fp4/flex · fp4
Inference by DeepInfra
138.2
From 318 benchmarks. Plenty of data behind it.
121 ±9
Coding IQ 121.0, 95% interval 112–130. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.45cache read $0.10$2.001M
57.7
Good value (57.7). Cheaper per task than its benchmark results predict. Plenty of data behind it.
65 ±10
Chatty (65 ±10). Generates more tokens than models with similar coding results. Measured at max effort. A fair amount of data behind it.
157.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GLM 5.3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
Inference by Digitalocean
138.2
From 318 benchmarks. Plenty of data behind it.
121 ±9
Coding IQ 121.0, 95% interval 112–130. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.91cache read $0.169$2.861M
57.7
Good value (57.7). Cheaper per task than its benchmark results predict. Plenty of data behind it.
65 ±10
Chatty (65 ±10). Generates more tokens than models with similar coding results. Measured at max effort. A fair amount of data behind it.
157.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GLM 5.3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
nvfp4
Inference by Doubleword
138.2
From 318 benchmarks. Plenty of data behind it.
121 ±9
Coding IQ 121.0, 95% interval 112–130. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.70cache read $0.14$2.201M
57.7
Good value (57.7). Cheaper per task than its benchmark results predict. Plenty of data behind it.
65 ±10
Chatty (65 ±10). Generates more tokens than models with similar coding results. Measured at max effort. A fair amount of data behind it.
157.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GLM 5.3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
Inference by Featherless AI
138.2
From 318 benchmarks. Plenty of data behind it.
121 ±9
Coding IQ 121.0, 95% interval 112–130. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.84cache read $0.156$2.64262K
57.7
Good value (57.7). Cheaper per task than its benchmark results predict. Plenty of data behind it.
65 ±10
Chatty (65 ±10). Generates more tokens than models with similar coding results. Measured at max effort. A fair amount of data behind it.
157.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GLM 5.3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
Inference by Fireworks AI
138.2
From 318 benchmarks. Plenty of data behind it.
121 ±9
Coding IQ 121.0, 95% interval 112–130. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$1.40cache read $0.26$4.401M
57.7
Good value (57.7). Cheaper per task than its benchmark results predict. Plenty of data behind it.
65 ±10
Chatty (65 ±10). Generates more tokens than models with similar coding results. Measured at max effort. A fair amount of data behind it.
157.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GLM 5.3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
fast
Inference by Fireworks AI
138.2
From 318 benchmarks. Plenty of data behind it.
121 ±9
Coding IQ 121.0, 95% interval 112–130. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$2.10cache read $0.39$6.601M
57.7
Good value (57.7). Cheaper per task than its benchmark results predict. Plenty of data behind it.
65 ±10
Chatty (65 ±10). Generates more tokens than models with similar coding results. Measured at max effort. A fair amount of data behind it.
157.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GLM 5.3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
spot
Inference by Fireworks AI
138.2
From 318 benchmarks. Plenty of data behind it.
121 ±9
Coding IQ 121.0, 95% interval 112–130. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.70cache read $0.13$2.201M
57.7
Good value (57.7). Cheaper per task than its benchmark results predict. Plenty of data behind it.
65 ±10
Chatty (65 ±10). Generates more tokens than models with similar coding results. Measured at max effort. A fair amount of data behind it.
157.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GLM 5.3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
Inference by Friendli AI
138.2
From 318 benchmarks. Plenty of data behind it.
121 ±9
Coding IQ 121.0, 95% interval 112–130. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$1.26cache read $0.234$3.961M
57.7
Good value (57.7). Cheaper per task than its benchmark results predict. Plenty of data behind it.
65 ±10
Chatty (65 ±10). Generates more tokens than models with similar coding results. Measured at max effort. A fair amount of data behind it.
157.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GLM 5.3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
fp4
Inference by Inceptron
138.2
From 318 benchmarks. Plenty of data behind it.
121 ±9
Coding IQ 121.0, 95% interval 112–130. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$1.05cache read $0.1171$3.501M
57.7
Good value (57.7). Cheaper per task than its benchmark results predict. Plenty of data behind it.
65 ±10
Chatty (65 ±10). Generates more tokens than models with similar coding results. Measured at max effort. A fair amount of data behind it.
157.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GLM 5.3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
fp4
Inference by Inferencenet
138.2
From 318 benchmarks. Plenty of data behind it.
121 ±9
Coding IQ 121.0, 95% interval 112–130. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.084cache read $0.079$3.601M
57.7
Good value (57.7). Cheaper per task than its benchmark results predict. Plenty of data behind it.
65 ±10
Chatty (65 ±10). Generates more tokens than models with similar coding results. Measured at max effort. A fair amount of data behind it.
157.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GLM 5.3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
us · fp4
Inference by Inferencenet
138.2
From 318 benchmarks. Plenty of data behind it.
121 ±9
Coding IQ 121.0, 95% interval 112–130. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.084cache read $0.079$3.601M
57.7
Good value (57.7). Cheaper per task than its benchmark results predict. Plenty of data behind it.
65 ±10
Chatty (65 ±10). Generates more tokens than models with similar coding results. Measured at max effort. A fair amount of data behind it.
157.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GLM 5.3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
fp4
Inference by Makora
138.2
From 318 benchmarks. Plenty of data behind it.
121 ±9
Coding IQ 121.0, 95% interval 112–130. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$1.50cache read $0.26$4.401M
57.7
Good value (57.7). Cheaper per task than its benchmark results predict. Plenty of data behind it.
65 ±10
Chatty (65 ±10). Generates more tokens than models with similar coding results. Measured at max effort. A fair amount of data behind it.
157.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GLM 5.3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
nvfp4/eu · nvfp4
Inference by Mistral AI
138.2
From 318 benchmarks. Plenty of data behind it.
121 ±9
Coding IQ 121.0, 95% interval 112–130. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$1.54cache read $0.154$4.841M
57.7
Good value (57.7). Cheaper per task than its benchmark results predict. Plenty of data behind it.
65 ±10
Chatty (65 ±10). Generates more tokens than models with similar coding results. Measured at max effort. A fair amount of data behind it.
157.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GLM 5.3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
zdr · nvfp4
Inference by Mistral AI
138.2
From 318 benchmarks. Plenty of data behind it.
121 ±9
Coding IQ 121.0, 95% interval 112–130. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$1.40cache read $0.14$4.401M
57.7
Good value (57.7). Cheaper per task than its benchmark results predict. Plenty of data behind it.
65 ±10
Chatty (65 ±10). Generates more tokens than models with similar coding results. Measured at max effort. A fair amount of data behind it.
157.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GLM 5.3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
Inference by Modal
138.2
From 318 benchmarks. Plenty of data behind it.
121 ±9
Coding IQ 121.0, 95% interval 112–130. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$1.40cache read $0.26$4.401M
57.7
Good value (57.7). Cheaper per task than its benchmark results predict. Plenty of data behind it.
65 ±10
Chatty (65 ±10). Generates more tokens than models with similar coding results. Measured at max effort. A fair amount of data behind it.
157.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GLM 5.3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
fp4
Inference by Modelrun
138.2
From 318 benchmarks. Plenty of data behind it.
121 ±9
Coding IQ 121.0, 95% interval 112–130. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$2.10cache read $0.40$6.601M
57.7
Good value (57.7). Cheaper per task than its benchmark results predict. Plenty of data behind it.
65 ±10
Chatty (65 ±10). Generates more tokens than models with similar coding results. Measured at max effort. A fair amount of data behind it.
157.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GLM 5.3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
z-ai/glm-5.3:batch · batch+fp8 · fp8
Inference by Morph
138.2
From 318 benchmarks. Plenty of data behind it.
121 ±9
Coding IQ 121.0, 95% interval 112–130. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.281cache read $0.062$1.251M
57.7
Good value (57.7). Cheaper per task than its benchmark results predict. Plenty of data behind it.
65 ±10
Chatty (65 ±10). Generates more tokens than models with similar coding results. Measured at max effort. A fair amount of data behind it.
157.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GLM 5.3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
fp8
Inference by Morph
138.2
From 318 benchmarks. Plenty of data behind it.
121 ±9
Coding IQ 121.0, 95% interval 112–130. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$1.006cache read $0.204$4.6961M
57.7
Good value (57.7). Cheaper per task than its benchmark results predict. Plenty of data behind it.
65 ±10
Chatty (65 ±10). Generates more tokens than models with similar coding results. Measured at max effort. A fair amount of data behind it.
157.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GLM 5.3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
fp4
Inference by Nebius AI Studio
138.2
From 318 benchmarks. Plenty of data behind it.
121 ±9
Coding IQ 121.0, 95% interval 112–130. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$1.40cache read $0.26$4.401M
57.7
Good value (57.7). Cheaper per task than its benchmark results predict. Plenty of data behind it.
65 ±10
Chatty (65 ±10). Generates more tokens than models with similar coding results. Measured at max effort. A fair amount of data behind it.
157.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GLM 5.3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
fp8
Inference by Novita AI
138.2
From 318 benchmarks. Plenty of data behind it.
121 ±9
Coding IQ 121.0, 95% interval 112–130. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.63cache read $0.117$1.981M
57.7
Good value (57.7). Cheaper per task than its benchmark results predict. Plenty of data behind it.
65 ±10
Chatty (65 ±10). Generates more tokens than models with similar coding results. Measured at max effort. A fair amount of data behind it.
157.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GLM 5.3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
fast · fp4
Inference by Parasail
138.2
From 318 benchmarks. Plenty of data behind it.
121 ±9
Coding IQ 121.0, 95% interval 112–130. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$2.10cache read $0.26$6.601M
57.7
Good value (57.7). Cheaper per task than its benchmark results predict. Plenty of data behind it.
65 ±10
Chatty (65 ±10). Generates more tokens than models with similar coding results. Measured at max effort. A fair amount of data behind it.
157.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GLM 5.3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
fp4
Inference by Parasail
138.2
From 318 benchmarks. Plenty of data behind it.
121 ±9
Coding IQ 121.0, 95% interval 112–130. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$1.40cache read $0.26$4.401M
57.7
Good value (57.7). Cheaper per task than its benchmark results predict. Plenty of data behind it.
65 ±10
Chatty (65 ±10). Generates more tokens than models with similar coding results. Measured at max effort. A fair amount of data behind it.
157.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GLM 5.3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
fp4
Inference by Relace
138.2
From 318 benchmarks. Plenty of data behind it.
121 ±9
Coding IQ 121.0, 95% interval 112–130. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.0703$4.38431M
57.7
Good value (57.7). Cheaper per task than its benchmark results predict. Plenty of data behind it.
65 ±10
Chatty (65 ±10). Generates more tokens than models with similar coding results. Measured at max effort. A fair amount of data behind it.
157.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GLM 5.3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
Inference by River
138.2
From 318 benchmarks. Plenty of data behind it.
121 ±9
Coding IQ 121.0, 95% interval 112–130. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$1.12cache read $0.208$3.521M
57.7
Good value (57.7). Cheaper per task than its benchmark results predict. Plenty of data behind it.
65 ±10
Chatty (65 ±10). Generates more tokens than models with similar coding results. Measured at max effort. A fair amount of data behind it.
157.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GLM 5.3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
fp8
Inference by Sail Research
138.2
From 318 benchmarks. Plenty of data behind it.
121 ±9
Coding IQ 121.0, 95% interval 112–130. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$1.05cache read $0.168$4.201M
57.7
Good value (57.7). Cheaper per task than its benchmark results predict. Plenty of data behind it.
65 ±10
Chatty (65 ±10). Generates more tokens than models with similar coding results. Measured at max effort. A fair amount of data behind it.
157.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GLM 5.3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
us · fp8
Inference by Sail Research
138.2
From 318 benchmarks. Plenty of data behind it.
121 ±9
Coding IQ 121.0, 95% interval 112–130. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.75cache read $0.12$3.001M
57.7
Good value (57.7). Cheaper per task than its benchmark results predict. Plenty of data behind it.
65 ±10
Chatty (65 ±10). Generates more tokens than models with similar coding results. Measured at max effort. A fair amount of data behind it.
157.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GLM 5.3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
Inference by SambaNova
138.2
From 318 benchmarks. Plenty of data behind it.
121 ±9
Coding IQ 121.0, 95% interval 112–130. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$1.40cache read $0.07$4.401M
57.7
Good value (57.7). Cheaper per task than its benchmark results predict. Plenty of data behind it.
65 ±10
Chatty (65 ±10). Generates more tokens than models with similar coding results. Measured at max effort. A fair amount of data behind it.
157.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GLM 5.3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
fp8
Inference by SiliconFlow
138.2
From 318 benchmarks. Plenty of data behind it.
121 ±9
Coding IQ 121.0, 95% interval 112–130. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.70cache read $0.13$1.901M
57.7
Good value (57.7). Cheaper per task than its benchmark results predict. Plenty of data behind it.
65 ±10
Chatty (65 ±10). Generates more tokens than models with similar coding results. Measured at max effort. A fair amount of data behind it.
157.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GLM 5.3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
fp8
Inference by Tenstorrent
138.2
From 318 benchmarks. Plenty of data behind it.
121 ±9
Coding IQ 121.0, 95% interval 112–130. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.60cache read $0.10$2.501M
57.7
Good value (57.7). Cheaper per task than its benchmark results predict. Plenty of data behind it.
65 ±10
Chatty (65 ±10). Generates more tokens than models with similar coding results. Measured at max effort. A fair amount of data behind it.
157.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GLM 5.3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
Inference by Together AI
138.2
From 318 benchmarks. Plenty of data behind it.
121 ±9
Coding IQ 121.0, 95% interval 112–130. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$1.40cache read $0.26$4.401M
57.7
Good value (57.7). Cheaper per task than its benchmark results predict. Plenty of data behind it.
65 ±10
Chatty (65 ±10). Generates more tokens than models with similar coding results. Measured at max effort. A fair amount of data behind it.
157.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GLM 5.3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
z-ai/glm-5.3:batch · batch
Inference by Together AI
138.2
From 318 benchmarks. Plenty of data behind it.
121 ±9
Coding IQ 121.0, 95% interval 112–130. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.28cache read $0.061$1.2491M
57.7
Good value (57.7). Cheaper per task than its benchmark results predict. Plenty of data behind it.
65 ±10
Chatty (65 ±10). Generates more tokens than models with similar coding results. Measured at max effort. A fair amount of data behind it.
157.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GLM 5.3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
Inference by Venice
138.2
From 318 benchmarks. Plenty of data behind it.
121 ±9
Coding IQ 121.0, 95% interval 112–130. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$1.40cache read $0.26$4.401M
57.7
Good value (57.7). Cheaper per task than its benchmark results predict. Plenty of data behind it.
65 ±10
Chatty (65 ±10). Generates more tokens than models with similar coding results. Measured at max effort. A fair amount of data behind it.
157.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GLM 5.3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
fp4
Inference by Wafer
138.2
From 318 benchmarks. Plenty of data behind it.
121 ±9
Coding IQ 121.0, 95% interval 112–130. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.179cache read $0.178$1.401M
57.7
Good value (57.7). Cheaper per task than its benchmark results predict. Plenty of data behind it.
65 ±10
Chatty (65 ±10). Generates more tokens than models with similar coding results. Measured at max effort. A fair amount of data behind it.
157.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GLM 5.3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
us · fp4
Inference by Wafer
138.2
From 318 benchmarks. Plenty of data behind it.
121 ±9
Coding IQ 121.0, 95% interval 112–130. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.0699cache read $0.0689$1.401M
57.7
Good value (57.7). Cheaper per task than its benchmark results predict. Plenty of data behind it.
65 ±10
Chatty (65 ±10). Generates more tokens than models with similar coding results. Measured at max effort. A fair amount of data behind it.
157.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GLM 5.3
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
fp8
Inference by Z.ai
138.2
From 318 benchmarks. Plenty of data behind it.
121 ±9
Coding IQ 121.0, 95% interval 112–130. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$1.40cache read $0.26$4.401M
57.7
Good value (57.7). Cheaper per task than its benchmark results predict. Plenty of data behind it.
65 ±10
Chatty (65 ±10). Generates more tokens than models with similar coding results. Measured at max effort. A fair amount of data behind it.
157.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Grok 4.6
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
138.0
From 220 benchmarks. Plenty of data behind it.
120 ±11
Coding IQ 120.1, 95% interval 109–131. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$2.00cache read $0.50$6.00500K
56.0
Good value (56.0). Cheaper per task than its benchmark results predict. Plenty of data behind it.
47 ±10
Typical (47 ±10). Generates about as many tokens as models with similar coding results. Measured at high effort. A fair amount of data behind it.
157.8
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Grok 4.6
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
x-ai/grok-4.6:batch · batch
138.0
From 220 benchmarks. Plenty of data behind it.
120 ±11
Coding IQ 120.1, 95% interval 109–131. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$1.60cache read $0.40$4.80500K
56.0
Good value (56.0). Cheaper per task than its benchmark results predict. Plenty of data behind it.
47 ±10
Typical (47 ±10). Generates about as many tokens as models with similar coding results. Measured at high effort. A fair amount of data behind it.
157.8
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Grok 4.6
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Inference by xAI
138.0
From 220 benchmarks. Plenty of data behind it.
120 ±11
Coding IQ 120.1, 95% interval 109–131. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$2.00cache read $0.50$6.00500K
56.0
Good value (56.0). Cheaper per task than its benchmark results predict. Plenty of data behind it.
47 ±10
Typical (47 ±10). Generates about as many tokens as models with similar coding results. Measured at high effort. A fair amount of data behind it.
157.8
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Grok 4.6
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
x-ai/grok-4.6:batch · batch
Inference by xAI
138.0
From 220 benchmarks. Plenty of data behind it.
120 ±11
Coding IQ 120.1, 95% interval 109–131. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$1.60cache read $0.40$4.80500K
56.0
Good value (56.0). Cheaper per task than its benchmark results predict. Plenty of data behind it.
47 ±10
Typical (47 ±10). Generates about as many tokens as models with similar coding results. Measured at high effort. A fair amount of data behind it.
157.8
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Grok 4.6
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
priority
Inference by xAI
138.0
From 220 benchmarks. Plenty of data behind it.
120 ±11
Coding IQ 120.1, 95% interval 109–131. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$4.00cache read $1.00$12.00500K
56.0
Good value (56.0). Cheaper per task than its benchmark results predict. Plenty of data behind it.
47 ±10
Typical (47 ±10). Generates about as many tokens as models with similar coding results. Measured at high effort. A fair amount of data behind it.
157.8
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Grok 4.6
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
zdr
Inference by xAI
138.0
From 220 benchmarks. Plenty of data behind it.
120 ±11
Coding IQ 120.1, 95% interval 109–131. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$2.00cache read $0.50$6.00500K
56.0
Good value (56.0). Cheaper per task than its benchmark results predict. Plenty of data behind it.
47 ±10
Typical (47 ±10). Generates about as many tokens as models with similar coding results. Measured at high effort. A fair amount of data behind it.
157.8
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Grok 4.6
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
zdr/priority
Inference by xAI
138.0
From 220 benchmarks. Plenty of data behind it.
120 ±11
Coding IQ 120.1, 95% interval 109–131. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$4.00cache read $1.00$12.00500K
56.0
Good value (56.0). Cheaper per task than its benchmark results predict. Plenty of data behind it.
47 ±10
Typical (47 ±10). Generates about as many tokens as models with similar coding results. Measured at high effort. A fair amount of data behind it.
157.8
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
DeepSeek V4.1 Flash
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
137.7
From 179 benchmarks. Plenty of data behind it.
125 ±16
Coding IQ 124.7, 95% interval 109–140. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.0498cache read $0.001$0.19921M
68.1
Good value (68.1). Cheaper per task than its benchmark results predict. Plenty of data behind it.
60 ±9
Chatty (60 ±9). Generates more tokens than models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
157
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Grok 4.7
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
137.7
From 249 benchmarks. Plenty of data behind it.
117 ±7
Coding IQ 117.4, 95% interval 111–124. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$2.00cache read $0.50$6.00500K
41.2
Poor value (41.2). Costs more per task than its benchmark results predict. Plenty of data behind it.
55 ±14
Typical (55 ±14). Generates about as many tokens as models with similar coding results. Measured at max effort. Little data behind it, so treat it as a rough estimate.
156.4
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Grok 4.7
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
x-ai/grok-4.7:batch · batch
137.7
From 249 benchmarks. Plenty of data behind it.
117 ±7
Coding IQ 117.4, 95% interval 111–124. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$1.60cache read $0.40$4.80500K
41.2
Poor value (41.2). Costs more per task than its benchmark results predict. Plenty of data behind it.
55 ±14
Typical (55 ±14). Generates about as many tokens as models with similar coding results. Measured at max effort. Little data behind it, so treat it as a rough estimate.
156.4
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Grok 4.7
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Inference by xAI
137.7
From 249 benchmarks. Plenty of data behind it.
117 ±7
Coding IQ 117.4, 95% interval 111–124. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$2.00cache read $0.50$6.00500K
41.2
Poor value (41.2). Costs more per task than its benchmark results predict. Plenty of data behind it.
55 ±14
Typical (55 ±14). Generates about as many tokens as models with similar coding results. Measured at max effort. Little data behind it, so treat it as a rough estimate.
156.4
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Grok 4.7
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
x-ai/grok-4.7:batch · batch
Inference by xAI
137.7
From 249 benchmarks. Plenty of data behind it.
117 ±7
Coding IQ 117.4, 95% interval 111–124. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$1.60cache read $0.40$4.80500K
41.2
Poor value (41.2). Costs more per task than its benchmark results predict. Plenty of data behind it.
55 ±14
Typical (55 ±14). Generates about as many tokens as models with similar coding results. Measured at max effort. Little data behind it, so treat it as a rough estimate.
156.4
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Grok 4.7
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
priority
Inference by xAI
137.7
From 249 benchmarks. Plenty of data behind it.
117 ±7
Coding IQ 117.4, 95% interval 111–124. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$4.00cache read $1.00$12.00500K
41.2
Poor value (41.2). Costs more per task than its benchmark results predict. Plenty of data behind it.
55 ±14
Typical (55 ±14). Generates about as many tokens as models with similar coding results. Measured at max effort. Little data behind it, so treat it as a rough estimate.
156.4
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Grok 4.7
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
zdr
Inference by xAI
137.7
From 249 benchmarks. Plenty of data behind it.
117 ±7
Coding IQ 117.4, 95% interval 111–124. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$2.00cache read $0.50$6.00500K
41.2
Poor value (41.2). Costs more per task than its benchmark results predict. Plenty of data behind it.
55 ±14
Typical (55 ±14). Generates about as many tokens as models with similar coding results. Measured at max effort. Little data behind it, so treat it as a rough estimate.
156.4
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Grok 4.7
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
zdr/priority
Inference by xAI
137.7
From 249 benchmarks. Plenty of data behind it.
117 ±7
Coding IQ 117.4, 95% interval 111–124. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$4.00cache read $1.00$12.00500K
41.2
Poor value (41.2). Costs more per task than its benchmark results predict. Plenty of data behind it.
55 ±14
Typical (55 ±14). Generates about as many tokens as models with similar coding results. Measured at max effort. Little data behind it, so treat it as a rough estimate.
156.4
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Muse Ember
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
New
Released 9 Nov 2026
137.6
From 32 benchmarks. A fair amount of data behind it.
≈ 120 ±12
Coding IQ 119.7, 95% interval 107–132. Little data behind it, so treat it as a rough estimate. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.10cache read $0.004$0.50cache write $0.1251M
81.8
Good value (81.8). Cheaper per task than its benchmark results predict. Plenty of data behind it.
55 ±9
Typical (55 ±9). Generates about as many tokens as models with similar coding results. Measured at max effort. Little data behind it, so treat it as a rough estimate.
Not measured
Muse Ember
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
New
Released 9 Nov 2026
global
Inference by Google
137.6
From 32 benchmarks. A fair amount of data behind it.
≈ 120 ±12
Coding IQ 119.7, 95% interval 107–132. Little data behind it, so treat it as a rough estimate. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.10cache read $0.004$0.501M
81.8
Good value (81.8). Cheaper per task than its benchmark results predict. Plenty of data behind it.
55 ±9
Typical (55 ±9). Generates about as many tokens as models with similar coding results. Measured at max effort. Little data behind it, so treat it as a rough estimate.
Not measured
Muse Ember
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
New
Released 9 Nov 2026
Inference by Meta
137.6
From 32 benchmarks. A fair amount of data behind it.
≈ 120 ±12
Coding IQ 119.7, 95% interval 107–132. Little data behind it, so treat it as a rough estimate. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.10cache read $0.004$0.50cache write $0.1251M
81.8
Good value (81.8). Cheaper per task than its benchmark results predict. Plenty of data behind it.
55 ±9
Typical (55 ±9). Generates about as many tokens as models with similar coding results. Measured at max effort. Little data behind it, so treat it as a rough estimate.
Not measured
Qwen3.8 Omni Flash
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
137.6
From 33 benchmarks. A fair amount of data behind it.
Not enough data yet$0.15cache read $0.016$0.471MNot enough data yetNot measuredNot measured
Qwen3.8 Omni Flash
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Inference by Alibaba Cloud
137.6
From 33 benchmarks. A fair amount of data behind it.
Not enough data yet$0.15cache read $0.016$0.47984KNot enough data yetNot measuredNot measured
Hy 4
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
137.4
From 46 benchmarks. Plenty of data behind it.
≈ 123 ±12
Coding IQ 123.4, 95% interval 111–136. Little data behind it, so treat it as a rough estimate. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.417cache read $0.021$1.25051M
56.3
Good value (56.3). Cheaper per task than its benchmark results predict. Plenty of data behind it.
61 ±9
Chatty (61 ±9). Generates more tokens than models with similar coding results. Measured at max effort. Little data behind it, so treat it as a rough estimate.
Not measured
Hy 4
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
fp4
Inference by DeepInfra
137.4
From 46 benchmarks. Plenty of data behind it.
≈ 123 ±12
Coding IQ 123.4, 95% interval 111–136. Little data behind it, so treat it as a rough estimate. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.834cache read $0.042$2.5011M
56.3
Good value (56.3). Cheaper per task than its benchmark results predict. Plenty of data behind it.
61 ±9
Chatty (61 ±9). Generates more tokens than models with similar coding results. Measured at max effort. Little data behind it, so treat it as a rough estimate.
Not measured
Hy 4
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
fp4
Inference by Tencent
137.4
From 46 benchmarks. Plenty of data behind it.
≈ 123 ±12
Coding IQ 123.4, 95% interval 111–136. Little data behind it, so treat it as a rough estimate. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.417cache read $0.021$1.25051M
56.3
Good value (56.3). Cheaper per task than its benchmark results predict. Plenty of data behind it.
61 ±9
Chatty (61 ±9). Generates more tokens than models with similar coding results. Measured at max effort. Little data behind it, so treat it as a rough estimate.
Not measured
Hy 4
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Inference by Together AI
137.4
From 46 benchmarks. Plenty of data behind it.
≈ 123 ±12
Coding IQ 123.4, 95% interval 111–136. Little data behind it, so treat it as a rough estimate. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.834cache read $0.042$2.5011M
56.3
Good value (56.3). Cheaper per task than its benchmark results predict. Plenty of data behind it.
61 ±9
Chatty (61 ±9). Generates more tokens than models with similar coding results. Measured at max effort. Little data behind it, so treat it as a rough estimate.
Not measured
Muse Spark 1.2
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Newer version
Newer versionMuse Spark 1.3We compared release dates and version names in our catalogue.
meta/muse-spark-1.2-contributor
137.3
From 199 benchmarks. Plenty of data behind it.
116 ±6
Coding IQ 116.1, 95% interval 110–122. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.10cache read $0.002$0.201M
62.0
Good value (62.0). Cheaper per task than its benchmark results predict. Plenty of data behind it.
66 ±9
Chatty (66 ±9). Generates more tokens than models with similar coding results. Measured at max effort. Little data behind it, so treat it as a rough estimate.
155
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Muse Spark 1.2
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Newer version
Newer versionMuse Spark 1.3We compared release dates and version names in our catalogue.
meta/muse-spark-1.2-contributor
Inference by Meta
137.3
From 199 benchmarks. Plenty of data behind it.
116 ±6
Coding IQ 116.1, 95% interval 110–122. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.10cache read $0.002$0.201M
62.0
Good value (62.0). Cheaper per task than its benchmark results predict. Plenty of data behind it.
66 ±9
Chatty (66 ±9). Generates more tokens than models with similar coding results. Measured at max effort. Little data behind it, so treat it as a rough estimate.
155
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Claude Opus 4.6
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Newer version
Newer versionClaude Opus 5.5We compared release dates and version names in our catalogue.
anthropic/claude-opus-4.6-fast · fast
136.9
From 466 benchmarks. Plenty of data behind it.
118 ±5
Coding IQ 117.9, 95% interval 113–123. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$30.00cache read $3.00$150.00cache write $37.50Not published by the provider
47.2
Poor value (47.2). Costs more per task than its benchmark results predict. Plenty of data behind it.
41 ±15
Terse (41 ±15). Generates fewer tokens than models with similar coding results. Measured at default effort. A fair amount of data behind it.
156.3
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Claude Opus 4.6
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Newer version
Newer versionClaude Opus 5.5We compared release dates and version names in our catalogue.
anthropic/claude-4.6-opus-20260205 · 20260205
136.9
From 466 benchmarks. Plenty of data behind it.
118 ±5
Coding IQ 117.9, 95% interval 113–123. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$5.275$26.375Not published by the provider
47.2
Poor value (47.2). Costs more per task than its benchmark results predict. Plenty of data behind it.
41 ±15
Terse (41 ±15). Generates fewer tokens than models with similar coding results. Measured at default effort. A fair amount of data behind it.
156.3
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Claude Opus 4.6
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Newer version
Newer versionClaude Opus 5.5We compared release dates and version names in our catalogue.
136.9
From 466 benchmarks. Plenty of data behind it.
118 ±5
Coding IQ 117.9, 95% interval 113–123. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$5.00cache read $0.50$25.00cache write $6.251M
47.2
Poor value (47.2). Costs more per task than its benchmark results predict. Plenty of data behind it.
41 ±15
Terse (41 ±15). Generates fewer tokens than models with similar coding results. Measured at default effort. A fair amount of data behind it.
156.3
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Claude Opus 4.6
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Newer version
Newer versionClaude Opus 5.5We compared release dates and version names in our catalogue.
anthropic/claude-opus-4.6:batch · batch
136.9
From 466 benchmarks. Plenty of data behind it.
118 ±5
Coding IQ 117.9, 95% interval 113–123. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$2.50cache read $0.25$12.50cache write $3.1251M
47.2
Poor value (47.2). Costs more per task than its benchmark results predict. Plenty of data behind it.
41 ±15
Terse (41 ±15). Generates fewer tokens than models with similar coding results. Measured at default effort. A fair amount of data behind it.
156.3
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Claude Opus 4.6
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Newer version
Newer versionClaude Opus 5.5We compared release dates and version names in our catalogue.
Inference by Amazon Web Services
136.9
From 466 benchmarks. Plenty of data behind it.
118 ±5
Coding IQ 117.9, 95% interval 113–123. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$5.00cache read $0.50$25.00cache write $6.251M
47.2
Poor value (47.2). Costs more per task than its benchmark results predict. Plenty of data behind it.
41 ±15
Terse (41 ±15). Generates fewer tokens than models with similar coding results. Measured at default effort. A fair amount of data behind it.
156.3
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Claude Opus 4.6
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Newer version
Newer versionClaude Opus 5.5We compared release dates and version names in our catalogue.
Inference by Anthropic
136.9
From 466 benchmarks. Plenty of data behind it.
118 ±5
Coding IQ 117.9, 95% interval 113–123. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$5.00cache read $0.50$25.00cache write $6.251M
47.2
Poor value (47.2). Costs more per task than its benchmark results predict. Plenty of data behind it.
41 ±15
Terse (41 ±15). Generates fewer tokens than models with similar coding results. Measured at default effort. A fair amount of data behind it.
156.3
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Claude Opus 4.6
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Newer version
Newer versionClaude Opus 5.5We compared release dates and version names in our catalogue.
global
Inference by Microsoft Azure
136.9
From 466 benchmarks. Plenty of data behind it.
118 ±5
Coding IQ 117.9, 95% interval 113–123. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$5.00cache read $0.50$25.00cache write $6.251M
47.2
Poor value (47.2). Costs more per task than its benchmark results predict. Plenty of data behind it.
41 ±15
Terse (41 ±15). Generates fewer tokens than models with similar coding results. Measured at default effort. A fair amount of data behind it.
156.3
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Claude Opus 4.6
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Newer version
Newer versionClaude Opus 5.5We compared release dates and version names in our catalogue.
Inference by Claude Platform On AWS
136.9
From 466 benchmarks. Plenty of data behind it.
118 ±5
Coding IQ 117.9, 95% interval 113–123. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$5.00cache read $0.50$25.00cache write $6.251M
47.2
Poor value (47.2). Costs more per task than its benchmark results predict. Plenty of data behind it.
41 ±15
Terse (41 ±15). Generates fewer tokens than models with similar coding results. Measured at default effort. A fair amount of data behind it.
156.3
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Claude Opus 4.6
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Newer version
Newer versionClaude Opus 5.5We compared release dates and version names in our catalogue.
global
Inference by Google
136.9
From 466 benchmarks. Plenty of data behind it.
118 ±5
Coding IQ 117.9, 95% interval 113–123. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$5.00cache read $0.50$25.00cache write $6.251M
47.2
Poor value (47.2). Costs more per task than its benchmark results predict. Plenty of data behind it.
41 ±15
Terse (41 ±15). Generates fewer tokens than models with similar coding results. Measured at default effort. A fair amount of data behind it.
156.3
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
DeepSeek V4 Pro 0813
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
136.8
From 147 benchmarks. Plenty of data behind it.
117 ±7
Coding IQ 117.1, 95% interval 110–124. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$1.32cache read $0.044$3.961M
52.5
Fair value (52.5). Costs per task about what its benchmark results predict. Plenty of data behind it.
58 ±9
Chatty (58 ±9). Generates more tokens than models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
156
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
DeepSeek V4 Pro 0813
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
Inference by Alibaba Cloud
136.8
From 147 benchmarks. Plenty of data behind it.
117 ±7
Coding IQ 117.1, 95% interval 110–124. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$1.122cache read $0.1122$3.3661M
52.5
Fair value (52.5). Costs per task about what its benchmark results predict. Plenty of data behind it.
58 ±9
Chatty (58 ±9). Generates more tokens than models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
156
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
DeepSeek V4 Pro 0813
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
us-east-1
Inference by Alibaba Cloud
136.8
From 147 benchmarks. Plenty of data behind it.
117 ±7
Coding IQ 117.1, 95% interval 110–124. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$1.32cache read $0.132$3.961M
52.5
Fair value (52.5). Costs per task about what its benchmark results predict. Plenty of data behind it.
58 ±9
Chatty (58 ±9). Generates more tokens than models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
156
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
DeepSeek V4 Pro 0813
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
fp8
Inference by Baidu
136.8
From 147 benchmarks. Plenty of data behind it.
117 ±7
Coding IQ 117.1, 95% interval 110–124. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.132cache read $0.0042$0.3961M
52.5
Fair value (52.5). Costs per task about what its benchmark results predict. Plenty of data behind it.
58 ±9
Chatty (58 ±9). Generates more tokens than models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
156
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
DeepSeek V4 Pro 0813
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
fp4
Inference by Baseten
136.8
From 147 benchmarks. Plenty of data behind it.
117 ±7
Coding IQ 117.1, 95% interval 110–124. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$1.32cache read $0.132$3.961M
52.5
Fair value (52.5). Costs per task about what its benchmark results predict. Plenty of data behind it.
58 ±9
Chatty (58 ±9). Generates more tokens than models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
156
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
DeepSeek V4 Pro 0813
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
fp8
Inference by Coreweave
136.8
From 147 benchmarks. Plenty of data behind it.
117 ±7
Coding IQ 117.1, 95% interval 110–124. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$1.31cache read $0.044$3.961M
52.5
Fair value (52.5). Costs per task about what its benchmark results predict. Plenty of data behind it.
58 ±9
Chatty (58 ±9). Generates more tokens than models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
156
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
DeepSeek V4 Pro 0813
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
fp8
Inference by DeepInfra
136.8
From 147 benchmarks. Plenty of data behind it.
117 ±7
Coding IQ 117.1, 95% interval 110–124. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$1.30cache read $0.10$2.601M
52.5
Fair value (52.5). Costs per task about what its benchmark results predict. Plenty of data behind it.
58 ±9
Chatty (58 ±9). Generates more tokens than models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
156
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
DeepSeek V4 Pro 0813
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
fp8/flex · fp8
Inference by DeepInfra
136.8
From 147 benchmarks. Plenty of data behind it.
117 ±7
Coding IQ 117.1, 95% interval 110–124. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$1.04cache read $0.08$2.081M
52.5
Fair value (52.5). Costs per task about what its benchmark results predict. Plenty of data behind it.
58 ±9
Chatty (58 ±9). Generates more tokens than models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
156
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
DeepSeek V4 Pro 0813
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
fp8/priority · fp8
Inference by DeepInfra
136.8
From 147 benchmarks. Plenty of data behind it.
117 ±7
Coding IQ 117.1, 95% interval 110–124. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$1.95cache read $0.15$3.901M
52.5
Fair value (52.5). Costs per task about what its benchmark results predict. Plenty of data behind it.
58 ±9
Chatty (58 ±9). Generates more tokens than models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
156
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
DeepSeek V4 Pro 0813
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
Inference by DeepSeek
136.8
From 147 benchmarks. Plenty of data behind it.
117 ±7
Coding IQ 117.1, 95% interval 110–124. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.66cache read $0.022$1.981M
52.5
Fair value (52.5). Costs per task about what its benchmark results predict. Plenty of data behind it.
58 ±9
Chatty (58 ±9). Generates more tokens than models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
156
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
DeepSeek V4 Pro 0813
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
Inference by Digitalocean
136.8
From 147 benchmarks. Plenty of data behind it.
117 ±7
Coding IQ 117.1, 95% interval 110–124. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$1.32cache read $0.044$3.961M
52.5
Fair value (52.5). Costs per task about what its benchmark results predict. Plenty of data behind it.
58 ±9
Chatty (58 ±9). Generates more tokens than models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
156
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
DeepSeek V4 Pro 0813
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
fp8
Inference by Nebius AI Studio
136.8
From 147 benchmarks. Plenty of data behind it.
117 ±7
Coding IQ 117.1, 95% interval 110–124. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$1.32cache read $0.04$3.961M
52.5
Fair value (52.5). Costs per task about what its benchmark results predict. Plenty of data behind it.
58 ±9
Chatty (58 ±9). Generates more tokens than models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
156
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
DeepSeek V4 Pro 0813
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
fp8
Inference by Novita AI
136.8
From 147 benchmarks. Plenty of data behind it.
117 ±7
Coding IQ 117.1, 95% interval 110–124. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$1.254cache read $0.0418$3.7621M
52.5
Fair value (52.5). Costs per task about what its benchmark results predict. Plenty of data behind it.
58 ±9
Chatty (58 ±9). Generates more tokens than models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
156
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
DeepSeek V4 Pro 0813
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
fast · fp8
Inference by Parasail
136.8
From 147 benchmarks. Plenty of data behind it.
117 ±7
Coding IQ 117.1, 95% interval 110–124. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$1.98cache read $0.066$5.941M
52.5
Fair value (52.5). Costs per task about what its benchmark results predict. Plenty of data behind it.
58 ±9
Chatty (58 ±9). Generates more tokens than models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
156
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
DeepSeek V4 Pro 0813
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
fp8
Inference by Parasail
136.8
From 147 benchmarks. Plenty of data behind it.
117 ±7
Coding IQ 117.1, 95% interval 110–124. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$1.32cache read $0.044$3.961M
52.5
Fair value (52.5). Costs per task about what its benchmark results predict. Plenty of data behind it.
58 ±9
Chatty (58 ±9). Generates more tokens than models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
156
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
DeepSeek V4 Pro 0813
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
fp4
Inference by Relace
136.8
From 147 benchmarks. Plenty of data behind it.
117 ±7
Coding IQ 117.1, 95% interval 110–124. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.0999$8.001M
52.5
Fair value (52.5). Costs per task about what its benchmark results predict. Plenty of data behind it.
58 ±9
Chatty (58 ±9). Generates more tokens than models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
156
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
DeepSeek V4 Pro 0813
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
fp8
Inference by SiliconFlow
136.8
From 147 benchmarks. Plenty of data behind it.
117 ±7
Coding IQ 117.1, 95% interval 110–124. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$1.32cache read $0.044$3.961M
52.5
Fair value (52.5). Costs per task about what its benchmark results predict. Plenty of data behind it.
58 ±9
Chatty (58 ±9). Generates more tokens than models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
156
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
DeepSeek V4 Pro 0813
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
Inference by Streamlake
136.8
From 147 benchmarks. Plenty of data behind it.
117 ±7
Coding IQ 117.1, 95% interval 110–124. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.66cache read $0.022$1.981M
52.5
Fair value (52.5). Costs per task about what its benchmark results predict. Plenty of data behind it.
58 ±9
Chatty (58 ±9). Generates more tokens than models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
156
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
DeepSeek V4 Pro 0813
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
Inference by Together AI
136.8
From 147 benchmarks. Plenty of data behind it.
117 ±7
Coding IQ 117.1, 95% interval 110–124. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$1.32cache read $0.13$3.961M
52.5
Fair value (52.5). Costs per task about what its benchmark results predict. Plenty of data behind it.
58 ±9
Chatty (58 ±9). Generates more tokens than models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
156
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
DeepSeek V4 Pro 0813
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
Inference by Venice
136.8
From 147 benchmarks. Plenty of data behind it.
117 ±7
Coding IQ 117.1, 95% interval 110–124. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$1.65cache read $0.165$4.951M
52.5
Fair value (52.5). Costs per task about what its benchmark results predict. Plenty of data behind it.
58 ±9
Chatty (58 ±9). Generates more tokens than models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
156
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
DeepSeek V4 Pro 0813
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
mxfp4
Inference by Wafer
136.8
From 147 benchmarks. Plenty of data behind it.
117 ±7
Coding IQ 117.1, 95% interval 110–124. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.285cache read $0.284$4.301M
52.5
Fair value (52.5). Costs per task about what its benchmark results predict. Plenty of data behind it.
58 ±9
Chatty (58 ±9). Generates more tokens than models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
156
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
DeepSeek V4 Pro 0813
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
us · mxfp4
Inference by Wafer
136.8
From 147 benchmarks. Plenty of data behind it.
117 ±7
Coding IQ 117.1, 95% interval 110–124. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.111cache read $0.097$4.441M
52.5
Fair value (52.5). Costs per task about what its benchmark results predict. Plenty of data behind it.
58 ±9
Chatty (58 ±9). Generates more tokens than models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
156
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GPT 5.4
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Newer version
Newer versionGPT 6We compared the version names in our catalogue.
136.8
From 489 benchmarks. Plenty of data behind it.
112 ±8
Coding IQ 112.0, 95% interval 104–120. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$2.50cache read $0.25$15.001.1M
60.1
Good value (60.1). Cheaper per task than its benchmark results predict. Plenty of data behind it.
28 ±8
Terse (28 ±8). Generates fewer tokens than models with similar coding results. Measured at default effort. A fair amount of data behind it.
157.9
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GPT 5.4
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Newer version
Newer versionGPT 6We compared the version names in our catalogue.
openai/gpt-5.4:batch · batch
136.8
From 489 benchmarks. Plenty of data behind it.
112 ±8
Coding IQ 112.0, 95% interval 104–120. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$1.25cache read $0.125$7.501.1M
60.1
Good value (60.1). Cheaper per task than its benchmark results predict. Plenty of data behind it.
28 ±8
Terse (28 ±8). Generates fewer tokens than models with similar coding results. Measured at default effort. A fair amount of data behind it.
157.9
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GPT 5.4
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Newer version
Newer versionGPT 6We compared the version names in our catalogue.
Inference by Microsoft Azure
136.8
From 489 benchmarks. Plenty of data behind it.
112 ±8
Coding IQ 112.0, 95% interval 104–120. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$2.50cache read $0.25$15.00922K
60.1
Good value (60.1). Cheaper per task than its benchmark results predict. Plenty of data behind it.
28 ±8
Terse (28 ±8). Generates fewer tokens than models with similar coding results. Measured at default effort. A fair amount of data behind it.
157.9
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GPT 5.4
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Newer version
Newer versionGPT 6We compared the version names in our catalogue.
Inference by OpenAI
136.8
From 489 benchmarks. Plenty of data behind it.
112 ±8
Coding IQ 112.0, 95% interval 104–120. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$2.50cache read $0.25$15.00922K
60.1
Good value (60.1). Cheaper per task than its benchmark results predict. Plenty of data behind it.
28 ±8
Terse (28 ±8). Generates fewer tokens than models with similar coding results. Measured at default effort. A fair amount of data behind it.
157.9
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GPT 5.4
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Newer version
Newer versionGPT 6We compared the version names in our catalogue.
fast
Inference by OpenAI
136.8
From 489 benchmarks. Plenty of data behind it.
112 ±8
Coding IQ 112.0, 95% interval 104–120. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$5.00cache read $0.50$30.00922K
60.1
Good value (60.1). Cheaper per task than its benchmark results predict. Plenty of data behind it.
28 ±8
Terse (28 ±8). Generates fewer tokens than models with similar coding results. Measured at default effort. A fair amount of data behind it.
157.9
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GPT 5.4
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Newer version
Newer versionGPT 6We compared the version names in our catalogue.
flex
Inference by OpenAI
136.8
From 489 benchmarks. Plenty of data behind it.
112 ±8
Coding IQ 112.0, 95% interval 104–120. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$1.25cache read $0.125$7.50922K
60.1
Good value (60.1). Cheaper per task than its benchmark results predict. Plenty of data behind it.
28 ±8
Terse (28 ±8). Generates fewer tokens than models with similar coding results. Measured at default effort. A fair amount of data behind it.
157.9
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Inkling 2 Small
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
New
Released 5 Nov 2026
136.8
From 97 benchmarks. Plenty of data behind it.
117 ±8
Coding IQ 116.9, 95% interval 109–125. A fair amount of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.30cache read $0.03$1.001M
63.2
Good value (63.2). Cheaper per task than its benchmark results predict. Plenty of data behind it.
Not measuredNot measured
Inkling 2 Small
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
New
Released 5 Nov 2026
Inference by Cloudflare Workers AI
136.8
From 97 benchmarks. Plenty of data behind it.
117 ±8
Coding IQ 116.9, 95% interval 109–125. A fair amount of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.30cache read $0.03$1.001M
63.2
Good value (63.2). Cheaper per task than its benchmark results predict. Plenty of data behind it.
Not measuredNot measured
Inkling 2 Small
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
New
Released 5 Nov 2026
fp4
Inference by DeepInfra
136.8
From 97 benchmarks. Plenty of data behind it.
117 ±8
Coding IQ 116.9, 95% interval 109–125. A fair amount of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.285cache read $0.0285$0.951M
63.2
Good value (63.2). Cheaper per task than its benchmark results predict. Plenty of data behind it.
Not measuredNot measured
Inkling 2 Small
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
New
Released 5 Nov 2026
nvfp4
Inference by Modal
136.8
From 97 benchmarks. Plenty of data behind it.
117 ±8
Coding IQ 116.9, 95% interval 109–125. A fair amount of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.30cache read $0.03$1.001M
63.2
Good value (63.2). Cheaper per task than its benchmark results predict. Plenty of data behind it.
Not measuredNot measured
Inkling 2 Small
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
New
Released 5 Nov 2026
fp4
Inference by Nebius AI Studio
136.8
From 97 benchmarks. Plenty of data behind it.
117 ±8
Coding IQ 116.9, 95% interval 109–125. A fair amount of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.30cache read $0.06$1.001M
63.2
Good value (63.2). Cheaper per task than its benchmark results predict. Plenty of data behind it.
Not measuredNot measured
Inkling 2 Small
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
New
Released 5 Nov 2026
fp4
Inference by Primeintellect
136.8
From 97 benchmarks. Plenty of data behind it.
117 ±8
Coding IQ 116.9, 95% interval 109–125. A fair amount of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.15cache read $0.03$0.501M
63.2
Good value (63.2). Cheaper per task than its benchmark results predict. Plenty of data behind it.
Not measuredNot measured
Inkling 2 Small
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
New
Released 5 Nov 2026
Inference by Together AI
136.8
From 97 benchmarks. Plenty of data behind it.
117 ±8
Coding IQ 116.9, 95% interval 109–125. A fair amount of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.30cache read $0.03$1.001M
63.2
Good value (63.2). Cheaper per task than its benchmark results predict. Plenty of data behind it.
Not measuredNot measured
MiMo V2.6 Pro
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
136.7
From 133 benchmarks. Plenty of data behind it.
123 ±9
Coding IQ 123.4, 95% interval 115–132. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.435cache read $0.0036$0.871M
74.5
Good value (74.5). Cheaper per task than its benchmark results predict. Plenty of data behind it.
62 ±9
Chatty (62 ±9). Generates more tokens than models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
Not measured
MiMo V2.6 Pro
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
xiaomi/mimo-v2.6-pro-ultraspeed · ultraspeed
136.7
From 133 benchmarks. Plenty of data behind it.
123 ±9
Coding IQ 123.4, 95% interval 115–132. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$4.35cache read $0.036$8.701M
74.5
Good value (74.5). Cheaper per task than its benchmark results predict. Plenty of data behind it.
62 ±9
Chatty (62 ±9). Generates more tokens than models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
Not measured
MiMo V2.6 Pro
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
fp8
Inference by DeepInfra
136.7
From 133 benchmarks. Plenty of data behind it.
123 ±9
Coding IQ 123.4, 95% interval 115–132. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.387cache read $0.0032$0.7831M
74.5
Good value (74.5). Cheaper per task than its benchmark results predict. Plenty of data behind it.
62 ±9
Chatty (62 ±9). Generates more tokens than models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
Not measured
MiMo V2.6 Pro
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
fp8
Inference by PrimaLabs
136.7
From 133 benchmarks. Plenty of data behind it.
123 ±9
Coding IQ 123.4, 95% interval 115–132. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$2.64cache read $0.024$5.281M
74.5
Good value (74.5). Cheaper per task than its benchmark results predict. Plenty of data behind it.
62 ±9
Chatty (62 ±9). Generates more tokens than models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
Not measured
MiMo V2.6 Pro
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
Inference by Streamlake
136.7
From 133 benchmarks. Plenty of data behind it.
123 ±9
Coding IQ 123.4, 95% interval 115–132. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.435cache read $0.0036$0.871M
74.5
Good value (74.5). Cheaper per task than its benchmark results predict. Plenty of data behind it.
62 ±9
Chatty (62 ±9). Generates more tokens than models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
Not measured
MiMo V2.6 Pro
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
fp8
Inference by Xiaomi
136.7
From 133 benchmarks. Plenty of data behind it.
123 ±9
Coding IQ 123.4, 95% interval 115–132. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.435cache read $0.0036$0.871M
74.5
Good value (74.5). Cheaper per task than its benchmark results predict. Plenty of data behind it.
62 ±9
Chatty (62 ±9). Generates more tokens than models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
Not measured
MiMo V2.6 Pro
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
xiaomi/mimo-v2.6-pro-ultraspeed · ultraspeed+fp8 · fp8
Inference by Xiaomi
136.7
From 133 benchmarks. Plenty of data behind it.
123 ±9
Coding IQ 123.4, 95% interval 115–132. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$4.35cache read $0.036$8.701M
74.5
Good value (74.5). Cheaper per task than its benchmark results predict. Plenty of data behind it.
62 ±9
Chatty (62 ±9). Generates more tokens than models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
Not measured
DeepSeek V4 Flash Vision Exp
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
136.6
From 32 benchmarks. A fair amount of data behind it.
≈ 110 ±19
Coding IQ 109.6, 95% interval 90–129. Little data behind it, so treat it as a rough estimate. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.2156cache read $0.0069$0.64681M
63.0
Good value (63.0). Cheaper per task than its benchmark results predict. Little data behind it, so treat it as a rough estimate.
Not measuredNot measured
DeepSeek V4 Flash Vision Exp
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
fp8
Inference by DeepInfra
136.6
From 32 benchmarks. A fair amount of data behind it.
≈ 110 ±19
Coding IQ 109.6, 95% interval 90–129. Little data behind it, so treat it as a rough estimate. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.2156cache read $0.0069$0.64681M
63.0
Good value (63.0). Cheaper per task than its benchmark results predict. Little data behind it, so treat it as a rough estimate.
Not measuredNot measured
DeepSeek V4 Flash Vision Exp
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
fp8/flex · fp8
Inference by DeepInfra
136.6
From 32 benchmarks. A fair amount of data behind it.
≈ 110 ±19
Coding IQ 109.6, 95% interval 90–129. Little data behind it, so treat it as a rough estimate. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.1725cache read $0.0055$0.51741M
63.0
Good value (63.0). Cheaper per task than its benchmark results predict. Little data behind it, so treat it as a rough estimate.
Not measuredNot measured
DeepSeek V4 Flash Vision Exp
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
fp8/priority · fp8
Inference by DeepInfra
136.6
From 32 benchmarks. A fair amount of data behind it.
≈ 110 ±19
Coding IQ 109.6, 95% interval 90–129. Little data behind it, so treat it as a rough estimate. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.3234cache read $0.0103$0.97021M
63.0
Good value (63.0). Cheaper per task than its benchmark results predict. Little data behind it, so treat it as a rough estimate.
Not measuredNot measured
DeepSeek V4 Flash Vision Exp
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
fp8
Inference by SiliconFlow
136.6
From 32 benchmarks. A fair amount of data behind it.
≈ 110 ±19
Coding IQ 109.6, 95% interval 90–129. Little data behind it, so treat it as a rough estimate. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.44cache read $0.028$1.321M
63.0
Good value (63.0). Cheaper per task than its benchmark results predict. Little data behind it, so treat it as a rough estimate.
Not measuredNot measured
Grok 4.5
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
136.6
From 312 benchmarks. Plenty of data behind it.
117 ±8
Coding IQ 117.4, 95% interval 109–125. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$2.00cache read $0.30$6.00500K
56.9
Good value (56.9). Cheaper per task than its benchmark results predict. Plenty of data behind it.
37 ±9
Terse (37 ±9). Generates fewer tokens than models with similar coding results. Measured at high effort. A fair amount of data behind it.
155
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Grok 4.5
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Inference by xAI
136.6
From 312 benchmarks. Plenty of data behind it.
117 ±8
Coding IQ 117.4, 95% interval 109–125. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$2.00cache read $0.30$6.00500K
56.9
Good value (56.9). Cheaper per task than its benchmark results predict. Plenty of data behind it.
37 ±9
Terse (37 ±9). Generates fewer tokens than models with similar coding results. Measured at high effort. A fair amount of data behind it.
155
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Grok 4.5
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
priority
Inference by xAI
136.6
From 312 benchmarks. Plenty of data behind it.
117 ±8
Coding IQ 117.4, 95% interval 109–125. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$4.00cache read $0.60$12.00500K
56.9
Good value (56.9). Cheaper per task than its benchmark results predict. Plenty of data behind it.
37 ±9
Terse (37 ±9). Generates fewer tokens than models with similar coding results. Measured at high effort. A fair amount of data behind it.
155
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Grok 4.5
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
zdr
Inference by xAI
136.6
From 312 benchmarks. Plenty of data behind it.
117 ±8
Coding IQ 117.4, 95% interval 109–125. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$2.00cache read $0.30$6.00500K
56.9
Good value (56.9). Cheaper per task than its benchmark results predict. Plenty of data behind it.
37 ±9
Terse (37 ±9). Generates fewer tokens than models with similar coding results. Measured at high effort. A fair amount of data behind it.
155
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Grok 4.5
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
zdr/priority
Inference by xAI
136.6
From 312 benchmarks. Plenty of data behind it.
117 ±8
Coding IQ 117.4, 95% interval 109–125. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$4.00cache read $0.60$12.00500K
56.9
Good value (56.9). Cheaper per task than its benchmark results predict. Plenty of data behind it.
37 ±9
Terse (37 ±9). Generates fewer tokens than models with similar coding results. Measured at high effort. A fair amount of data behind it.
155
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Hy4 preview
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
136.6
From 95 benchmarks. Plenty of data behind it.
118 ±16
Coding IQ 118.2, 95% interval 102–135. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.834cache read $0.042$2.5011M
56.4
Good value (56.4). Cheaper per task than its benchmark results predict. Plenty of data behind it.
57 ±9
Chatty (57 ±9). Generates more tokens than models with similar coding results. Measured at max effort. Little data behind it, so treat it as a rough estimate.
Not measured
Hy4 preview
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
fp4
Inference by DeepInfra
136.6
From 95 benchmarks. Plenty of data behind it.
118 ±16
Coding IQ 118.2, 95% interval 102–135. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.834cache read $0.042$2.5011M
56.4
Good value (56.4). Cheaper per task than its benchmark results predict. Plenty of data behind it.
57 ±9
Chatty (57 ±9). Generates more tokens than models with similar coding results. Measured at max effort. Little data behind it, so treat it as a rough estimate.
Not measured
Hy4 preview
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
fp8
Inference by Novita AI
136.6
From 95 benchmarks. Plenty of data behind it.
118 ±16
Coding IQ 118.2, 95% interval 102–135. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.834cache read $0.042$2.5011M
56.4
Good value (56.4). Cheaper per task than its benchmark results predict. Plenty of data behind it.
57 ±9
Chatty (57 ±9). Generates more tokens than models with similar coding results. Measured at max effort. Little data behind it, so treat it as a rough estimate.
Not measured
Hy4 preview
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
fp8
Inference by SiliconFlow
136.6
From 95 benchmarks. Plenty of data behind it.
118 ±16
Coding IQ 118.2, 95% interval 102–135. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.834cache read $0.042$2.5011M
56.4
Good value (56.4). Cheaper per task than its benchmark results predict. Plenty of data behind it.
57 ±9
Chatty (57 ±9). Generates more tokens than models with similar coding results. Measured at max effort. Little data behind it, so treat it as a rough estimate.
Not measured
Hy4 preview
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
fp8
Inference by Tencent
136.6
From 95 benchmarks. Plenty of data behind it.
118 ±16
Coding IQ 118.2, 95% interval 102–135. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.834cache read $0.042$2.5011M
56.4
Good value (56.4). Cheaper per task than its benchmark results predict. Plenty of data behind it.
57 ±9
Chatty (57 ±9). Generates more tokens than models with similar coding results. Measured at max effort. Little data behind it, so treat it as a rough estimate.
Not measured
Hy4 preview
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
fp4
Inference by Venice
136.6
From 95 benchmarks. Plenty of data behind it.
118 ±16
Coding IQ 118.2, 95% interval 102–135. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$1.0425cache read $0.0525$3.12621M
56.4
Good value (56.4). Cheaper per task than its benchmark results predict. Plenty of data behind it.
57 ±9
Chatty (57 ±9). Generates more tokens than models with similar coding results. Measured at max effort. Little data behind it, so treat it as a rough estimate.
Not measured
Muse Spark 1.1
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Newer version
Newer versionMuse Spark 1.3We compared release dates and version names in our catalogue.
136.6
From 203 benchmarks. Plenty of data behind it.
115 ±5
Coding IQ 115.0, 95% interval 110–120. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$1.25cache read $0.15$4.251M
67.4
Good value (67.4). Cheaper per task than its benchmark results predict. Plenty of data behind it.
58 ±16
Chatty (58 ±16). Generates more tokens than models with similar coding results. Measured at max effort. A fair amount of data behind it.
154.9
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Muse Spark 1.1
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Newer version
Newer versionMuse Spark 1.3We compared release dates and version names in our catalogue.
Inference by Meta
136.6
From 203 benchmarks. Plenty of data behind it.
115 ±5
Coding IQ 115.0, 95% interval 110–120. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$1.25cache read $0.15$4.251M
67.4
Good value (67.4). Cheaper per task than its benchmark results predict. Plenty of data behind it.
58 ±16
Chatty (58 ±16). Generates more tokens than models with similar coding results. Measured at max effort. A fair amount of data behind it.
154.9
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Gemini 3 Pro Preview
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Newer version
Newer versionGemini 3.1 Pro PreviewWe compared release dates and version names in our catalogue.
136.4
From 490 benchmarks. Plenty of data behind it.
111 ±7
Coding IQ 110.6, 95% interval 103–118. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$2.00$12.00Not published by the provider
54.0
Good value (54.0). Cheaper per task than its benchmark results predict. Plenty of data behind it.
43 ±9
Terse (43 ±9). Generates fewer tokens than models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
153
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Claude Sonnet 5
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Newer version
Newer versionClaude Sonnet 5.5We compared release dates and version names in our catalogue.
136.3
From 360 benchmarks. Plenty of data behind it.
115 ±5
Coding IQ 115.2, 95% interval 111–120. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$2.00cache read $0.20$10.00cache write $2.501M
37.4
Poor value (37.4). Costs more per task than its benchmark results predict. Plenty of data behind it.
57 ±9
Chatty (57 ±9). Generates more tokens than models with similar coding results. Measured at medium effort. Little data behind it, so treat it as a rough estimate.
156.9
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Claude Sonnet 5
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Newer version
Newer versionClaude Sonnet 5.5We compared release dates and version names in our catalogue.
anthropic/claude-sonnet-5:batch · batch
136.3
From 360 benchmarks. Plenty of data behind it.
115 ±5
Coding IQ 115.2, 95% interval 111–120. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$1.00cache read $0.10$5.00cache write $1.251M
37.4
Poor value (37.4). Costs more per task than its benchmark results predict. Plenty of data behind it.
57 ±9
Chatty (57 ±9). Generates more tokens than models with similar coding results. Measured at medium effort. Little data behind it, so treat it as a rough estimate.
156.9
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Claude Sonnet 5
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Newer version
Newer versionClaude Sonnet 5.5We compared release dates and version names in our catalogue.
global
Inference by Amazon Web Services
136.3
From 360 benchmarks. Plenty of data behind it.
115 ±5
Coding IQ 115.2, 95% interval 111–120. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$2.00cache read $0.20$10.00cache write $2.501M
37.4
Poor value (37.4). Costs more per task than its benchmark results predict. Plenty of data behind it.
57 ±9
Chatty (57 ±9). Generates more tokens than models with similar coding results. Measured at medium effort. Little data behind it, so treat it as a rough estimate.
156.9
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Claude Sonnet 5
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Newer version
Newer versionClaude Sonnet 5.5We compared release dates and version names in our catalogue.
Inference by Anthropic
136.3
From 360 benchmarks. Plenty of data behind it.
115 ±5
Coding IQ 115.2, 95% interval 111–120. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$2.00cache read $0.20$10.00cache write $2.501M
37.4
Poor value (37.4). Costs more per task than its benchmark results predict. Plenty of data behind it.
57 ±9
Chatty (57 ±9). Generates more tokens than models with similar coding results. Measured at medium effort. Little data behind it, so treat it as a rough estimate.
156.9
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Claude Sonnet 5
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Newer version
Newer versionClaude Sonnet 5.5We compared release dates and version names in our catalogue.
global
Inference by Microsoft Azure
136.3
From 360 benchmarks. Plenty of data behind it.
115 ±5
Coding IQ 115.2, 95% interval 111–120. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$2.00cache read $0.20$10.00cache write $2.501M
37.4
Poor value (37.4). Costs more per task than its benchmark results predict. Plenty of data behind it.
57 ±9
Chatty (57 ±9). Generates more tokens than models with similar coding results. Measured at medium effort. Little data behind it, so treat it as a rough estimate.
156.9
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Claude Sonnet 5
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Newer version
Newer versionClaude Sonnet 5.5We compared release dates and version names in our catalogue.
Inference by Claude Platform On AWS
136.3
From 360 benchmarks. Plenty of data behind it.
115 ±5
Coding IQ 115.2, 95% interval 111–120. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$2.00cache read $0.20$10.00cache write $2.501M
37.4
Poor value (37.4). Costs more per task than its benchmark results predict. Plenty of data behind it.
57 ±9
Chatty (57 ±9). Generates more tokens than models with similar coding results. Measured at medium effort. Little data behind it, so treat it as a rough estimate.
156.9
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Claude Sonnet 5
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Newer version
Newer versionClaude Sonnet 5.5We compared release dates and version names in our catalogue.
global
Inference by Google
136.3
From 360 benchmarks. Plenty of data behind it.
115 ±5
Coding IQ 115.2, 95% interval 111–120. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$2.00cache read $0.20$10.00cache write $2.501M
37.4
Poor value (37.4). Costs more per task than its benchmark results predict. Plenty of data behind it.
57 ±9
Chatty (57 ±9). Generates more tokens than models with similar coding results. Measured at medium effort. Little data behind it, so treat it as a rough estimate.
156.9
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API

Data credits: Epoch AI Benchmarking Hub, CC BY 4.0 (licence)

How this data works