Skip to content

Listings

All listings: NanoGPT

Every listing for the models on page 6 of 40 of NanoGPT, including each model's other versions and routes. Back to NanoGPT

Sold by NanoGPT

111 listings on this page sold by NanoGPT.

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
Solar Mini 4
🇰🇷
Maker headquartered in KR, according to the company's own website. The citation is on the model page.
125.4
From 12 benchmarks. A fair amount of data behind it.
Not enough data yet$0.05cache read $0.005$0.20524KNot enough data yetNot measuredNot measured
Solar Mini 4
🇰🇷
Maker headquartered in KR, according to the company's own website. The citation is on the model page.
upstage/solar-mini4:thinking · thinking
125.4
From 12 benchmarks. A fair amount of data behind it.
Not enough data yet$0.05cache read $0.005$0.20524KNot enough data yetNot measuredNot measured
Gemini 3.1 Flash Lite
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Newer version
Newer versionGemini 3.5 Flash LiteWe compared release dates and version names in our catalogue.
125.3
From 275 benchmarks. Plenty of data behind it.
97 ±10
Coding IQ 97.2, 95% interval 88–107. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.25cache read $0.025$1.501M
55.5
Good value (55.5). Cheaper per task than its benchmark results predict. Plenty of data behind it.
48 ±9
Typical (48 ±9). Generates about as many tokens as models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
145
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Gemini Omni 1.1 Flash
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Newer version
Newer versionGemini Omni 1.2 FlashWe compared the version names in our catalogue.
google/gemini-omni-flash/v1.1
≈ 125
From 8 benchmarks. Little data behind it, so treat it as a rough estimate.
Not enough data yet——Not published by the providerNot enough data yetNot measuredNot measured
DeepSeek V3.2
🇨🇳
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 versionDeepSeek V4We compared the version names in our catalogue.
125.0
From 247 benchmarks. Plenty of data behind it.
101 ±7
Coding IQ 101.2, 95% interval 94–108. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.26cache read $0.13$0.38163K
63.8
Good value (63.8). Cheaper per task than its benchmark results predict. Plenty of data behind it.
Not measured
146.3
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
DeepSeek V3.2
🇨🇳
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 versionDeepSeek V4We compared the version names in our catalogue.
tee
125.0
From 247 benchmarks. Plenty of data behind it.
101 ±7
Coding IQ 101.2, 95% interval 94–108. 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$1.00164K
63.8
Good value (63.8). Cheaper per task than its benchmark results predict. Plenty of data behind it.
Not measured
146.3
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
DeepSeek V3.2
🇨🇳
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 versionDeepSeek V4We compared the version names in our catalogue.
deepseek/deepseek-v3.2:thinking · thinking
125.0
From 247 benchmarks. Plenty of data behind it.
101 ±7
Coding IQ 101.2, 95% interval 94–108. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.26cache read $0.13$0.38163K
63.8
Good value (63.8). Cheaper per task than its benchmark results predict. Plenty of data behind it.
Not measured
146.3
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
DeepSeek V3.2
🇨🇳
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 versionDeepSeek V4We compared the version names in our catalogue.
fp8 · 🇨🇳 CN
Inference by Baidu
125.0
From 247 benchmarks. Plenty of data behind it.
101 ±7
Coding IQ 101.2, 95% interval 94–108. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.294cache read $0.0294$0.441Not published by the provider
63.8
Good value (63.8). Cheaper per task than its benchmark results predict. Plenty of data behind it.
Not measured
146.3
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
DeepSeek V3.2
🇨🇳
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 versionDeepSeek V4We compared the version names in our catalogue.
deepseek/deepseek-v3.2:thinking · thinking · fp8 · 🇨🇳 CN
Inference by Baidu
125.0
From 247 benchmarks. Plenty of data behind it.
101 ±7
Coding IQ 101.2, 95% interval 94–108. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.294cache read $0.0294$0.441Not published by the provider
63.8
Good value (63.8). Cheaper per task than its benchmark results predict. Plenty of data behind it.
Not measured
146.3
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
DeepSeek V3.2
🇨🇳
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 versionDeepSeek V4We compared the version names in our catalogue.
fp4 · 🇺🇸 US
Inference by DeepInfra
125.0
From 247 benchmarks. Plenty of data behind it.
101 ±7
Coding IQ 101.2, 95% interval 94–108. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.273cache read $0.1365$0.399Not published by the provider
63.8
Good value (63.8). Cheaper per task than its benchmark results predict. Plenty of data behind it.
Not measured
146.3
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
DeepSeek V3.2
🇨🇳
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 versionDeepSeek V4We compared the version names in our catalogue.
deepseek/deepseek-v3.2:thinking · thinking · fp4 · 🇺🇸 US
Inference by DeepInfra
125.0
From 247 benchmarks. Plenty of data behind it.
101 ±7
Coding IQ 101.2, 95% interval 94–108. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.273cache read $0.1365$0.399Not published by the provider
63.8
Good value (63.8). Cheaper per task than its benchmark results predict. Plenty of data behind it.
Not measured
146.3
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
DeepSeek V3.2
🇨🇳
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 versionDeepSeek V4We compared the version names in our catalogue.
fp8 · 🇸🇬 SG
Inference by SiliconFlow
125.0
From 247 benchmarks. Plenty of data behind it.
101 ±7
Coding IQ 101.2, 95% interval 94–108. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.272cache read $0.147$0.84Not published by the provider
63.8
Good value (63.8). Cheaper per task than its benchmark results predict. Plenty of data behind it.
Not measured
146.3
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
DeepSeek V3.2
🇨🇳
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 versionDeepSeek V4We compared the version names in our catalogue.
deepseek/deepseek-v3.2:thinking · thinking · fp8 · 🇸🇬 SG
Inference by SiliconFlow
125.0
From 247 benchmarks. Plenty of data behind it.
101 ±7
Coding IQ 101.2, 95% interval 94–108. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.272cache read $0.147$0.84Not published by the provider
63.8
Good value (63.8). Cheaper per task than its benchmark results predict. Plenty of data behind it.
Not measured
146.3
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Gemma 4 26B A4B
🇺🇸
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.
google/gemma-4-26b-a4b-it
124.9
From 216 benchmarks. Plenty of data behind it.
102 ±8
Coding IQ 101.7, 95% interval 94–110. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.0354cache read $0.0177$0.0943262K
56.4
Good value (56.4). Cheaper per task than its benchmark results predict. A fair amount of data behind it.
50 ±9
Typical (50 ±9). Generates about as many tokens as models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
141.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Gemma 4 26B A4B
🇺🇸
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.
google/gemma-4-26b-a4b-it:thinking · thinking
124.9
From 216 benchmarks. Plenty of data behind it.
102 ±8
Coding IQ 101.7, 95% interval 94–110. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.0675cache read $0.0375$0.225262K
56.4
Good value (56.4). Cheaper per task than its benchmark results predict. A fair amount of data behind it.
50 ±9
Typical (50 ±9). Generates about as many tokens as models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
141.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Gemma 4 26B A4B
🇺🇸
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.
google/gemma-4-26b-a4b-it · fp4 · GLB
Inference by Akash
124.9
From 216 benchmarks. Plenty of data behind it.
102 ±8
Coding IQ 101.7, 95% interval 94–110. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.0735cache read $0.0483$0.252Not published by the provider
56.4
Good value (56.4). Cheaper per task than its benchmark results predict. A fair amount of data behind it.
50 ±9
Typical (50 ±9). Generates about as many tokens as models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
141.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Gemma 4 26B A4B
🇺🇸
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.
google/gemma-4-26b-a4b-it:thinking · thinking · fp4 · GLB
Inference by Akash
124.9
From 216 benchmarks. Plenty of data behind it.
102 ±8
Coding IQ 101.7, 95% interval 94–110. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.0735cache read $0.0483$0.252Not published by the provider
56.4
Good value (56.4). Cheaper per task than its benchmark results predict. A fair amount of data behind it.
50 ±9
Typical (50 ±9). Generates about as many tokens as models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
141.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Gemma 4 26B A4B
🇺🇸
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.
google/gemma-4-26b-a4b-it · FP8 · 🇰🇷 KR
Inference by Aoru
124.9
From 216 benchmarks. Plenty of data behind it.
102 ±8
Coding IQ 101.7, 95% interval 94–110. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.126cache read $0.063$0.399Not published by the provider
56.4
Good value (56.4). Cheaper per task than its benchmark results predict. A fair amount of data behind it.
50 ±9
Typical (50 ±9). Generates about as many tokens as models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
141.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Gemma 4 26B A4B
🇺🇸
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.
google/gemma-4-26b-a4b-it · bf16 · 🇺🇸 US
Inference by Coreweave
124.9
From 216 benchmarks. Plenty of data behind it.
102 ±8
Coding IQ 101.7, 95% interval 94–110. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.105cache read $0.0525$0.315Not published by the provider
56.4
Good value (56.4). Cheaper per task than its benchmark results predict. A fair amount of data behind it.
50 ±9
Typical (50 ±9). Generates about as many tokens as models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
141.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Gemma 4 26B A4B
🇺🇸
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.
google/gemma-4-26b-a4b-it:thinking · thinking · bf16 · 🇺🇸 US
Inference by Coreweave
124.9
From 216 benchmarks. Plenty of data behind it.
102 ±8
Coding IQ 101.7, 95% interval 94–110. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.105cache read $0.0525$0.315Not published by the provider
56.4
Good value (56.4). Cheaper per task than its benchmark results predict. A fair amount of data behind it.
50 ±9
Typical (50 ±9). Generates about as many tokens as models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
141.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Gemma 4 26B A4B
🇺🇸
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.
google/gemma-4-26b-a4b-it · int8 · GLB
Inference by Darkbloom
124.9
From 216 benchmarks. Plenty of data behind it.
102 ±8
Coding IQ 101.7, 95% interval 94–110. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.0525$0.1995Not published by the provider
56.4
Good value (56.4). Cheaper per task than its benchmark results predict. A fair amount of data behind it.
50 ±9
Typical (50 ±9). Generates about as many tokens as models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
141.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Gemma 4 26B A4B
🇺🇸
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.
google/gemma-4-26b-a4b-it:thinking · thinking · int4 · GLB
Inference by Darkbloom
124.9
From 216 benchmarks. Plenty of data behind it.
102 ±8
Coding IQ 101.7, 95% interval 94–110. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.0525$0.1995Not published by the provider
56.4
Good value (56.4). Cheaper per task than its benchmark results predict. A fair amount of data behind it.
50 ±9
Typical (50 ±9). Generates about as many tokens as models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
141.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Gemma 4 26B A4B
🇺🇸
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.
google/gemma-4-26b-a4b-it · fp8 · 🇺🇸 US
Inference by DeepInfra
124.9
From 216 benchmarks. Plenty of data behind it.
102 ±8
Coding IQ 101.7, 95% interval 94–110. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.0735cache read $0.0105$0.357Not published by the provider
56.4
Good value (56.4). Cheaper per task than its benchmark results predict. A fair amount of data behind it.
50 ±9
Typical (50 ±9). Generates about as many tokens as models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
141.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Gemma 4 26B A4B
🇺🇸
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.
google/gemma-4-26b-a4b-it · 🇺🇸 US
Inference by Google
124.9
From 216 benchmarks. Plenty of data behind it.
102 ±8
Coding IQ 101.7, 95% interval 94–110. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.1575$0.63Not published by the provider
56.4
Good value (56.4). Cheaper per task than its benchmark results predict. A fair amount of data behind it.
50 ±9
Typical (50 ±9). Generates about as many tokens as models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
141.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Gemma 4 26B A4B
🇺🇸
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.
google/gemma-4-26b-a4b-it:thinking · thinking · 🇺🇸 US
Inference by Google
124.9
From 216 benchmarks. Plenty of data behind it.
102 ±8
Coding IQ 101.7, 95% interval 94–110. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.1575$0.63Not published by the provider
56.4
Good value (56.4). Cheaper per task than its benchmark results predict. A fair amount of data behind it.
50 ±9
Typical (50 ±9). Generates about as many tokens as models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
141.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Gemma 4 26B A4B
🇺🇸
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.
google/gemma-4-26b-a4b-it · bf16 · 🇺🇸 US
Inference by Ionet
124.9
From 216 benchmarks. Plenty of data behind it.
102 ±8
Coding IQ 101.7, 95% interval 94–110. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.0704cache read $0.041$0.3045Not published by the provider
56.4
Good value (56.4). Cheaper per task than its benchmark results predict. A fair amount of data behind it.
50 ±9
Typical (50 ±9). Generates about as many tokens as models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
141.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Gemma 4 26B A4B
🇺🇸
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.
google/gemma-4-26b-a4b-it:thinking · thinking · bf16 · 🇺🇸 US
Inference by Ionet
124.9
From 216 benchmarks. Plenty of data behind it.
102 ±8
Coding IQ 101.7, 95% interval 94–110. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.0704cache read $0.041$0.3045Not published by the provider
56.4
Good value (56.4). Cheaper per task than its benchmark results predict. A fair amount of data behind it.
50 ±9
Typical (50 ±9). Generates about as many tokens as models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
141.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Gemma 4 26B A4B
🇺🇸
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.
google/gemma-4-26b-a4b-it · int4 · GLB
Inference by Kitani
124.9
From 216 benchmarks. Plenty of data behind it.
102 ±8
Coding IQ 101.7, 95% interval 94–110. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.1134cache read $0.0567$0.378Not published by the provider
56.4
Good value (56.4). Cheaper per task than its benchmark results predict. A fair amount of data behind it.
50 ±9
Typical (50 ±9). Generates about as many tokens as models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
141.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Gemma 4 26B A4B
🇺🇸
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.
google/gemma-4-26b-a4b-it · bf16 · 🇪🇺 EU
Inference by Nextbit
124.9
From 216 benchmarks. Plenty of data behind it.
102 ±8
Coding IQ 101.7, 95% interval 94–110. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.0709cache read $0.0394$0.2363Not published by the provider
56.4
Good value (56.4). Cheaper per task than its benchmark results predict. A fair amount of data behind it.
50 ±9
Typical (50 ±9). Generates about as many tokens as models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
141.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Gemma 4 26B A4B
🇺🇸
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.
google/gemma-4-26b-a4b-it:thinking · thinking · bf16 · 🇪🇺 EU
Inference by Nextbit
124.9
From 216 benchmarks. Plenty of data behind it.
102 ±8
Coding IQ 101.7, 95% interval 94–110. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.0709cache read $0.0394$0.2363Not published by the provider
56.4
Good value (56.4). Cheaper per task than its benchmark results predict. A fair amount of data behind it.
50 ±9
Typical (50 ±9). Generates about as many tokens as models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
141.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Gemma 4 26B A4B
🇺🇸
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.
google/gemma-4-26b-a4b-it · bf16 · 🇺🇸 US
Inference by Novita AI
124.9
From 216 benchmarks. Plenty of data behind it.
102 ±8
Coding IQ 101.7, 95% interval 94–110. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.1365$0.42Not published by the provider
56.4
Good value (56.4). Cheaper per task than its benchmark results predict. A fair amount of data behind it.
50 ±9
Typical (50 ±9). Generates about as many tokens as models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
141.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Gemma 4 26B A4B
🇺🇸
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.
google/gemma-4-26b-a4b-it:thinking · thinking · bf16 · 🇺🇸 US
Inference by Novita AI
124.9
From 216 benchmarks. Plenty of data behind it.
102 ±8
Coding IQ 101.7, 95% interval 94–110. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.1365$0.42Not published by the provider
56.4
Good value (56.4). Cheaper per task than its benchmark results predict. A fair amount of data behind it.
50 ±9
Typical (50 ±9). Generates about as many tokens as models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
141.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Gemma 4 26B A4B
🇺🇸
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.
google/gemma-4-26b-a4b-it · bf16 · 🇺🇸 US
Inference by Parasail
124.9
From 216 benchmarks. Plenty of data behind it.
102 ±8
Coding IQ 101.7, 95% interval 94–110. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.1365cache read $0.0525$0.42Not published by the provider
56.4
Good value (56.4). Cheaper per task than its benchmark results predict. A fair amount of data behind it.
50 ±9
Typical (50 ±9). Generates about as many tokens as models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
141.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Gemma 4 26B A4B
🇺🇸
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.
google/gemma-4-26b-a4b-it:thinking · thinking · bf16 · 🇺🇸 US
Inference by Parasail
124.9
From 216 benchmarks. Plenty of data behind it.
102 ±8
Coding IQ 101.7, 95% interval 94–110. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.1365cache read $0.0525$0.42Not published by the provider
56.4
Good value (56.4). Cheaper per task than its benchmark results predict. A fair amount of data behind it.
50 ±9
Typical (50 ±9). Generates about as many tokens as models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
141.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Gemma 4 26B A4B
🇺🇸
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.
google/gemma-4-26b-a4b-it · bf16 · UNK
Inference by Primeintellect
124.9
From 216 benchmarks. Plenty of data behind it.
102 ±8
Coding IQ 101.7, 95% interval 94–110. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.105cache read $0.0525$0.315Not published by the provider
56.4
Good value (56.4). Cheaper per task than its benchmark results predict. A fair amount of data behind it.
50 ±9
Typical (50 ±9). Generates about as many tokens as models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
141.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Gemma 4 26B A4B
🇺🇸
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.
google/gemma-4-26b-a4b-it:thinking · thinking · bf16 · UNK
Inference by Primeintellect
124.9
From 216 benchmarks. Plenty of data behind it.
102 ±8
Coding IQ 101.7, 95% interval 94–110. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.105cache read $0.0525$0.315Not published by the provider
56.4
Good value (56.4). Cheaper per task than its benchmark results predict. A fair amount of data behind it.
50 ±9
Typical (50 ±9). Generates about as many tokens as models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
141.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Gemma 4 26B A4B
🇺🇸
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.
google/gemma-4-26b-a4b-it · fp8 · 🇸🇬 SG
Inference by SiliconFlow
124.9
From 216 benchmarks. Plenty of data behind it.
102 ±8
Coding IQ 101.7, 95% interval 94–110. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.147cache read $0.0525$0.42Not published by the provider
56.4
Good value (56.4). Cheaper per task than its benchmark results predict. A fair amount of data behind it.
50 ±9
Typical (50 ±9). Generates about as many tokens as models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
141.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Gemma 4 26B A4B
🇺🇸
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.
google/gemma-4-26b-a4b-it:thinking · thinking · fp8 · 🇸🇬 SG
Inference by SiliconFlow
124.9
From 216 benchmarks. Plenty of data behind it.
102 ±8
Coding IQ 101.7, 95% interval 94–110. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.147cache read $0.0525$0.42Not published by the provider
56.4
Good value (56.4). Cheaper per task than its benchmark results predict. A fair amount of data behind it.
50 ±9
Typical (50 ±9). Generates about as many tokens as models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
141.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Gemma 4 26B A4B
🇺🇸
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.
google/gemma-4-26b-a4b-it · fp8 · GLB
Inference by Uomi
124.9
From 216 benchmarks. Plenty of data behind it.
102 ±8
Coding IQ 101.7, 95% interval 94–110. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.1158$0.4079Not published by the provider
56.4
Good value (56.4). Cheaper per task than its benchmark results predict. A fair amount of data behind it.
50 ±9
Typical (50 ±9). Generates about as many tokens as models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
141.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Gemma 4 26B A4B
🇺🇸
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.
google/gemma-4-26b-a4b-it:thinking · thinking · fp8 · GLB
Inference by Uomi
124.9
From 216 benchmarks. Plenty of data behind it.
102 ±8
Coding IQ 101.7, 95% interval 94–110. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.1158$0.4079Not published by the provider
56.4
Good value (56.4). Cheaper per task than its benchmark results predict. A fair amount of data behind it.
50 ±9
Typical (50 ±9). Generates about as many tokens as models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
141.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GPT-5 mini
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Newer version
Newer versionGPT-5.4 miniWe compared release dates and version names in our catalogue.
124.6
From 388 benchmarks. Plenty of data behind it.
98 ±14
Coding IQ 98.3, 95% interval 85–112. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.25cache read $0.025$2.00400K
55.1
Good value (55.1). Cheaper per task than its benchmark results predict. Plenty of data behind it.
40 ±9
Terse (40 ±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.
146
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Ling 3.0 Flash VL
🇨🇳
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.
≈ 125
From 3 benchmarks. Little data behind it, so treat it as a rough estimate.
Not enough data yet$0.06cache read $0.012$0.18262K
65.3
Good value (65.3). Cheaper per task than its benchmark results predict. Little data behind it, so treat it as a rough estimate.
Not measuredNot measured
Qwen 3 235B A22B Thinking 2507
🇨🇳
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.
124.3
From 161 benchmarks. Plenty of data behind it.
100 ±8stale
Coding IQ 99.6, 95% interval 91–108. A fair amount of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ. Its newest result is over a year old.
$0.30cache read $0.15$0.50131K
40.6
Poor value (40.6). Costs more per task than its benchmark results predict. A fair amount of data behind it.
Not measuredNot measured
Qwen 3 235B A22B Thinking 2507
🇨🇳
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 · 🇺🇸 US
Inference by Venice
124.3
From 161 benchmarks. Plenty of data behind it.
100 ±8stale
Coding IQ 99.6, 95% interval 91–108. A fair amount of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ. Its newest result is over a year old.
$0.4725$3.675Not published by the provider
40.6
Poor value (40.6). Costs more per task than its benchmark results predict. A fair amount of data behind it.
Not measuredNot measured
Grok 4.3
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
124.2
From 205 benchmarks. Plenty of data behind it.
101 ±8
Coding IQ 101.2, 95% interval 94–109. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$1.25cache read $0.20$2.501M
39.3
Poor value (39.3). Costs more per task than its benchmark results predict. Plenty of data behind it.
48 ±9
Typical (48 ±9). Generates about as many tokens as models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
Not measured
Grok 4.3
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
🇺🇸 US
Inference by xAI
124.2
From 205 benchmarks. Plenty of data behind it.
101 ±8
Coding IQ 101.2, 95% interval 94–109. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$1.3125cache read $0.21$2.625Not published by the provider
39.3
Poor value (39.3). Costs more per task than its benchmark results predict. Plenty of data behind it.
48 ±9
Typical (48 ±9). Generates about as many tokens as models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
Not measured
Nemotron 3 Ultra
🇺🇸
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.
nvidia/nemotron-3-ultra-550b-a55b
124.1
From 172 benchmarks. Plenty of data behind it.
101 ±12
Coding IQ 100.9, 95% interval 89–112. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.50cache read $0.15$2.50512K
33.1
Poor value (33.1). Costs more per task than its benchmark results predict. A fair amount of data behind it.
53 ±9
Typical (53 ±9). Generates about as many tokens as models with similar coding results. Measured with reasoning off. Little data behind it, so treat it as a rough estimate.
147
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Nemotron 3 Ultra
🇺🇸
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.
nvidia/nemotron-3-ultra-550b-a55b:thinking · thinking
124.1
From 172 benchmarks. Plenty of data behind it.
101 ±12
Coding IQ 100.9, 95% interval 89–112. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.50cache read $0.15$2.501M
33.1
Poor value (33.1). Costs more per task than its benchmark results predict. A fair amount of data behind it.
53 ±9
Typical (53 ±9). Generates about as many tokens as models with similar coding results. Measured with reasoning off. Little data behind it, so treat it as a rough estimate.
147
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Nemotron 3 Ultra
🇺🇸
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.
nvidia/nemotron-3-ultra-550b-a55b · 🇺🇸 US
Inference by DeepInfra
124.1
From 172 benchmarks. Plenty of data behind it.
101 ±12
Coding IQ 100.9, 95% interval 89–112. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.525cache read $0.105$2.31Not published by the provider
33.1
Poor value (33.1). Costs more per task than its benchmark results predict. A fair amount of data behind it.
53 ±9
Typical (53 ±9). Generates about as many tokens as models with similar coding results. Measured with reasoning off. Little data behind it, so treat it as a rough estimate.
147
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Nemotron 3 Ultra
🇺🇸
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.
nvidia/nemotron-3-ultra-550b-a55b:thinking · thinking · 🇺🇸 US
Inference by DeepInfra
124.1
From 172 benchmarks. Plenty of data behind it.
101 ±12
Coding IQ 100.9, 95% interval 89–112. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.525cache read $0.105$2.31Not published by the provider
33.1
Poor value (33.1). Costs more per task than its benchmark results predict. A fair amount of data behind it.
53 ±9
Typical (53 ±9). Generates about as many tokens as models with similar coding results. Measured with reasoning off. Little data behind it, so treat it as a rough estimate.
147
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Nemotron 3 Ultra
🇺🇸
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.
nvidia/nemotron-3-ultra-550b-a55b · UNK
Inference by Ollama
124.1
From 172 benchmarks. Plenty of data behind it.
101 ±12
Coding IQ 100.9, 95% interval 89–112. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.105$3.15Not published by the provider
33.1
Poor value (33.1). Costs more per task than its benchmark results predict. A fair amount of data behind it.
53 ±9
Typical (53 ±9). Generates about as many tokens as models with similar coding results. Measured with reasoning off. Little data behind it, so treat it as a rough estimate.
147
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Nemotron 3 Ultra
🇺🇸
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.
nvidia/nemotron-3-ultra-550b-a55b:thinking · thinking · UNK
Inference by Ollama
124.1
From 172 benchmarks. Plenty of data behind it.
101 ±12
Coding IQ 100.9, 95% interval 89–112. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.105$3.15Not published by the provider
33.1
Poor value (33.1). Costs more per task than its benchmark results predict. A fair amount of data behind it.
53 ±9
Typical (53 ±9). Generates about as many tokens as models with similar coding results. Measured with reasoning off. Little data behind it, so treat it as a rough estimate.
147
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Nemotron 3 Ultra
🇺🇸
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.
nvidia/nemotron-3-ultra-550b-a55b · 🇺🇸 US
Inference by Together AI
124.1
From 172 benchmarks. Plenty of data behind it.
101 ±12
Coding IQ 100.9, 95% interval 89–112. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.63cache read $0.21$3.78Not published by the provider
33.1
Poor value (33.1). Costs more per task than its benchmark results predict. A fair amount of data behind it.
53 ±9
Typical (53 ±9). Generates about as many tokens as models with similar coding results. Measured with reasoning off. Little data behind it, so treat it as a rough estimate.
147
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Nemotron 3 Ultra
🇺🇸
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.
nvidia/nemotron-3-ultra-550b-a55b:thinking · thinking · 🇺🇸 US
Inference by Together AI
124.1
From 172 benchmarks. Plenty of data behind it.
101 ±12
Coding IQ 100.9, 95% interval 89–112. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.63cache read $0.21$3.78Not published by the provider
33.1
Poor value (33.1). Costs more per task than its benchmark results predict. A fair amount of data behind it.
53 ±9
Typical (53 ±9). Generates about as many tokens as models with similar coding results. Measured with reasoning off. Little data behind it, so treat it as a rough estimate.
147
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Nemotron 3 Ultra
🇺🇸
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.
nvidia/nemotron-3-ultra-550b-a55b · bf16 · GLB
Inference by Uomi
124.1
From 172 benchmarks. Plenty of data behind it.
101 ±12
Coding IQ 100.9, 95% interval 89–112. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.6891$3.4453Not published by the provider
33.1
Poor value (33.1). Costs more per task than its benchmark results predict. A fair amount of data behind it.
53 ±9
Typical (53 ±9). Generates about as many tokens as models with similar coding results. Measured with reasoning off. Little data behind it, so treat it as a rough estimate.
147
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Nemotron 3 Ultra
🇺🇸
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.
nvidia/nemotron-3-ultra-550b-a55b:thinking · thinking · bf16 · GLB
Inference by Uomi
124.1
From 172 benchmarks. Plenty of data behind it.
101 ±12
Coding IQ 100.9, 95% interval 89–112. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.6891$3.4453Not published by the provider
33.1
Poor value (33.1). Costs more per task than its benchmark results predict. A fair amount of data behind it.
53 ±9
Typical (53 ±9). Generates about as many tokens as models with similar coding results. Measured with reasoning off. Little data behind it, so treat it as a rough estimate.
147
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Qwen3 Embedding 4B
🇨🇳
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.
124.1
From 196 benchmarks. Plenty of data behind it.
Not enough data yet$0.02—33KNot enough data yetNot measuredNot measured
GPT-5.1 Codex Mini
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
123.8
From 24 benchmarks. Plenty of data behind it.
Not enough data yet$0.25cache read $0.025$2.00400K
62.1
Good value (62.1). Cheaper per task than its benchmark results predict. Little data behind it, so treat it as a rough estimate.
Not measuredNot measured
Kimi K2 Thinking
🇨🇳
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.
123.8
From 184 benchmarks. Plenty of data behind it.
100 ±6
Coding IQ 100.2, 95% interval 94–106. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.60cache read $0.15$2.50262K
41.6
Poor value (41.6). Costs more per task than its benchmark results predict. Plenty of data behind it.
50 ±18
Typical (50 ±18). Generates about as many tokens as models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
145.5
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Kimi K2 Thinking
🇨🇳
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.
bf16 · 🇺🇸 US
Inference by Novita AI
123.8
From 184 benchmarks. Plenty of data behind it.
100 ±6
Coding IQ 100.2, 95% interval 94–106. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.63cache read $0.1575$2.625Not published by the provider
41.6
Poor value (41.6). Costs more per task than its benchmark results predict. Plenty of data behind it.
50 ±18
Typical (50 ±18). Generates about as many tokens as models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
145.5
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
DeepSeek V3.2 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.
123.7
From 179 benchmarks. Plenty of data behind it.
98 ±7
Coding IQ 97.5, 95% interval 90–105. A fair amount of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.28cache read $0.14$0.42164K
68.3
Good value (68.3). Cheaper per task than its benchmark results predict. Plenty of data behind it.
Not measured
144.4
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
DeepSeek V3.2 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.
deepseek-ai/deepseek-v3.2-exp-thinking · thinking
123.7
From 179 benchmarks. Plenty of data behind it.
98 ±7
Coding IQ 97.5, 95% interval 90–105. A fair amount of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.28cache read $0.14$0.42164K
68.3
Good value (68.3). Cheaper per task than its benchmark results predict. Plenty of data behind it.
Not measured
144.4
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Gemini Embedding 001
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
123.5
From 68 benchmarks. Plenty of data behind it.
Not enough data yet$0.15—8KNot enough data yetNot measuredNot measured
Qwen3.6 27B
🇨🇳
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 versionQwen 3.8 27BWe compared release dates and version names in our catalogue.
123.4
From 160 benchmarks. Plenty of data behind it.
101 ±6
Coding IQ 100.9, 95% interval 95–107. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.203cache read $0.1015$2.24262K
39.6
Poor value (39.6). Costs more per task than its benchmark results predict. Plenty of data behind it.
Not measured
146.7
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Qwen3.6 27B
🇨🇳
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 versionQwen 3.8 27BWe compared release dates and version names in our catalogue.
tee
123.4
From 160 benchmarks. Plenty of data behind it.
101 ±6
Coding IQ 100.9, 95% interval 95–107. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.32cache read $0.03$3.25262K
39.6
Poor value (39.6). Costs more per task than its benchmark results predict. Plenty of data behind it.
Not measured
146.7
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Qwen3.6 27B
🇨🇳
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 versionQwen 3.8 27BWe compared release dates and version names in our catalogue.
qwen/qwen3.6-27b:thinking · thinking
123.4
From 160 benchmarks. Plenty of data behind it.
101 ±6
Coding IQ 100.9, 95% interval 95–107. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.203cache read $0.1015$2.24262K
39.6
Poor value (39.6). Costs more per task than its benchmark results predict. Plenty of data behind it.
Not measured
146.7
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Qwen3.6 27B
🇨🇳
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 versionQwen 3.8 27BWe compared release dates and version names in our catalogue.
🇸🇬 SG
Inference by Alibaba Cloud
123.4
From 160 benchmarks. Plenty of data behind it.
101 ±6
Coding IQ 100.9, 95% interval 95–107. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.63$3.78Not published by the provider
39.6
Poor value (39.6). Costs more per task than its benchmark results predict. Plenty of data behind it.
Not measured
146.7
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Qwen3.6 27B
🇨🇳
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 versionQwen 3.8 27BWe compared release dates and version names in our catalogue.
qwen/qwen3.6-27b:thinking · thinking · 🇸🇬 SG
Inference by Alibaba Cloud
123.4
From 160 benchmarks. Plenty of data behind it.
101 ±6
Coding IQ 100.9, 95% interval 95–107. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.63$3.78Not published by the provider
39.6
Poor value (39.6). Costs more per task than its benchmark results predict. Plenty of data behind it.
Not measured
146.7
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Qwen3.6 27B
🇨🇳
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 versionQwen 3.8 27BWe compared release dates and version names in our catalogue.
fp8 · 🇺🇸 US
Inference by Chutes
123.4
From 160 benchmarks. Plenty of data behind it.
101 ±6
Coding IQ 100.9, 95% interval 95–107. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.2625cache read $0.0263$2.10Not published by the provider
39.6
Poor value (39.6). Costs more per task than its benchmark results predict. Plenty of data behind it.
Not measured
146.7
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Qwen3.6 27B
🇨🇳
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 versionQwen 3.8 27BWe compared release dates and version names in our catalogue.
qwen/qwen3.6-27b:thinking · thinking · fp8 · 🇺🇸 US
Inference by Chutes
123.4
From 160 benchmarks. Plenty of data behind it.
101 ±6
Coding IQ 100.9, 95% interval 95–107. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.2625cache read $0.0263$2.10Not published by the provider
39.6
Poor value (39.6). Costs more per task than its benchmark results predict. Plenty of data behind it.
Not measured
146.7
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Qwen3.6 27B
🇨🇳
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 versionQwen 3.8 27BWe compared release dates and version names in our catalogue.
fp8 · 🇺🇸 US
Inference by DeepInfra
123.4
From 160 benchmarks. Plenty of data behind it.
101 ±6
Coding IQ 100.9, 95% interval 95–107. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.336$3.36Not published by the provider
39.6
Poor value (39.6). Costs more per task than its benchmark results predict. Plenty of data behind it.
Not measured
146.7
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Qwen3.6 27B
🇨🇳
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 versionQwen 3.8 27BWe compared release dates and version names in our catalogue.
qwen/qwen3.6-27b:thinking · thinking · fp8 · 🇺🇸 US
Inference by DeepInfra
123.4
From 160 benchmarks. Plenty of data behind it.
101 ±6
Coding IQ 100.9, 95% interval 95–107. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.336$3.36Not published by the provider
39.6
Poor value (39.6). Costs more per task than its benchmark results predict. Plenty of data behind it.
Not measured
146.7
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Qwen3.6 27B
🇨🇳
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 versionQwen 3.8 27BWe compared release dates and version names in our catalogue.
fp8 · 🇪🇺 EU
Inference by Greenference
123.4
From 160 benchmarks. Plenty of data behind it.
101 ±6
Coding IQ 100.9, 95% interval 95–107. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.1575cache read $0.0788$1.4175Not published by the provider
39.6
Poor value (39.6). Costs more per task than its benchmark results predict. Plenty of data behind it.
Not measured
146.7
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Qwen3.6 27B
🇨🇳
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 versionQwen 3.8 27BWe compared release dates and version names in our catalogue.
🇺🇸 US
Inference by Novita AI
123.4
From 160 benchmarks. Plenty of data behind it.
101 ±6
Coding IQ 100.9, 95% interval 95–107. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.63$3.78Not published by the provider
39.6
Poor value (39.6). Costs more per task than its benchmark results predict. Plenty of data behind it.
Not measured
146.7
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Qwen3.6 27B
🇨🇳
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 versionQwen 3.8 27BWe compared release dates and version names in our catalogue.
qwen/qwen3.6-27b:thinking · thinking · 🇺🇸 US
Inference by Novita AI
123.4
From 160 benchmarks. Plenty of data behind it.
101 ±6
Coding IQ 100.9, 95% interval 95–107. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.63$3.78Not published by the provider
39.6
Poor value (39.6). Costs more per task than its benchmark results predict. Plenty of data behind it.
Not measured
146.7
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Qwen3.6 27B
🇨🇳
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 versionQwen 3.8 27BWe compared release dates and version names in our catalogue.
UNK
Inference by Reka AI
123.4
From 160 benchmarks. Plenty of data behind it.
101 ±6
Coding IQ 100.9, 95% interval 95–107. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.203$2.24Not published by the provider
39.6
Poor value (39.6). Costs more per task than its benchmark results predict. Plenty of data behind it.
Not measured
146.7
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Qwen3.6 27B
🇨🇳
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 versionQwen 3.8 27BWe compared release dates and version names in our catalogue.
qwen/qwen3.6-27b:thinking · thinking · UNK
Inference by Reka AI
123.4
From 160 benchmarks. Plenty of data behind it.
101 ±6
Coding IQ 100.9, 95% interval 95–107. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.203$2.24Not published by the provider
39.6
Poor value (39.6). Costs more per task than its benchmark results predict. Plenty of data behind it.
Not measured
146.7
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Qwen3.6 27B
🇨🇳
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 versionQwen 3.8 27BWe compared release dates and version names in our catalogue.
fp8 · 🇸🇬 SG
Inference by SiliconFlow
123.4
From 160 benchmarks. Plenty of data behind it.
101 ±6
Coding IQ 100.9, 95% interval 95–107. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.315$3.36Not published by the provider
39.6
Poor value (39.6). Costs more per task than its benchmark results predict. Plenty of data behind it.
Not measured
146.7
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Qwen3.6 27B
🇨🇳
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 versionQwen 3.8 27BWe compared release dates and version names in our catalogue.
qwen/qwen3.6-27b:thinking · thinking · fp8 · 🇸🇬 SG
Inference by SiliconFlow
123.4
From 160 benchmarks. Plenty of data behind it.
101 ±6
Coding IQ 100.9, 95% interval 95–107. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.315$3.36Not published by the provider
39.6
Poor value (39.6). Costs more per task than its benchmark results predict. Plenty of data behind it.
Not measured
146.7
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Qwen3.6 27B
🇨🇳
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 versionQwen 3.8 27BWe compared release dates and version names in our catalogue.
GLB
Inference by Uomi
123.4
From 160 benchmarks. Plenty of data behind it.
101 ±6
Coding IQ 100.9, 95% interval 95–107. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.3548$3.4931Not published by the provider
39.6
Poor value (39.6). Costs more per task than its benchmark results predict. Plenty of data behind it.
Not measured
146.7
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Qwen3.6 27B
🇨🇳
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 versionQwen 3.8 27BWe compared release dates and version names in our catalogue.
qwen/qwen3.6-27b:thinking · thinking · GLB
Inference by Uomi
123.4
From 160 benchmarks. Plenty of data behind it.
101 ±6
Coding IQ 100.9, 95% interval 95–107. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.3548$3.4931Not published by the provider
39.6
Poor value (39.6). Costs more per task than its benchmark results predict. Plenty of data behind it.
Not measured
146.7
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Solar Pro 4
🇰🇷
Maker headquartered in KR, according to the company's own website. The citation is on the model page.
123.3
From 94 benchmarks. Plenty of data behind it.
102 ±8
Coding IQ 102.1, 95% interval 94–110. 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$0.60524KNot enough data yetNot measuredNot measured
Solar Pro 4
🇰🇷
Maker headquartered in KR, according to the company's own website. The citation is on the model page.
upstage/solar-pro4:thinking · thinking
123.3
From 94 benchmarks. Plenty of data behind it.
102 ±8
Coding IQ 102.1, 95% interval 94–110. 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$0.60524KNot enough data yetNot measuredNot measured
GLM 4.6
🇨🇳
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 versionGLM 5.3We compared release dates and version names in our catalogue.
123.1
From 174 benchmarks. Plenty of data behind it.
96 ±22
Coding IQ 96.0, 95% interval 74–118. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.55cache read $0.11$2.20200K
46.8
Poor value (46.8). Costs more per task than its benchmark results predict. Plenty of data behind it.
48 ±9
Typical (48 ±9). Generates about as many tokens as models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
Not measured
GLM 4.6
🇨🇳
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 versionGLM 5.3We compared release dates and version names in our catalogue.
z-ai/glm-4.6-original · original
123.1
From 174 benchmarks. Plenty of data behind it.
96 ±22
Coding IQ 96.0, 95% interval 74–118. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.55cache read $0.11$2.20256K
46.8
Poor value (46.8). Costs more per task than its benchmark results predict. Plenty of data behind it.
48 ±9
Typical (48 ±9). Generates about as many tokens as models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
Not measured
GLM 4.6
🇨🇳
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 versionGLM 5.3We compared release dates and version names in our catalogue.
z-ai/glm-4.6:thinking · thinking
123.1
From 174 benchmarks. Plenty of data behind it.
96 ±22
Coding IQ 96.0, 95% interval 74–118. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.55cache read $0.11$2.20200K
46.8
Poor value (46.8). Costs more per task than its benchmark results predict. Plenty of data behind it.
48 ±9
Typical (48 ±9). Generates about as many tokens as models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
Not measured
GLM 4.6
🇨🇳
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 versionGLM 5.3We compared release dates and version names in our catalogue.
fp4 · 🇺🇸 US
Inference by DeepInfra
123.1
From 174 benchmarks. Plenty of data behind it.
96 ±22
Coding IQ 96.0, 95% interval 74–118. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.525cache read $0.105$2.10Not published by the provider
46.8
Poor value (46.8). Costs more per task than its benchmark results predict. Plenty of data behind it.
48 ±9
Typical (48 ±9). Generates about as many tokens as models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
Not measured
GLM 4.6
🇨🇳
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 versionGLM 5.3We compared release dates and version names in our catalogue.
z-ai/glm-4.6-original · original · fp4 · 🇺🇸 US
Inference by DeepInfra
123.1
From 174 benchmarks. Plenty of data behind it.
96 ±22
Coding IQ 96.0, 95% interval 74–118. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.525cache read $0.105$2.10Not published by the provider
46.8
Poor value (46.8). Costs more per task than its benchmark results predict. Plenty of data behind it.
48 ±9
Typical (48 ±9). Generates about as many tokens as models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
Not measured
GLM 4.6
🇨🇳
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 versionGLM 5.3We compared release dates and version names in our catalogue.
z-ai/glm-4.6:thinking · thinking · fp4 · 🇺🇸 US
Inference by DeepInfra
123.1
From 174 benchmarks. Plenty of data behind it.
96 ±22
Coding IQ 96.0, 95% interval 74–118. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.525cache read $0.105$2.10Not published by the provider
46.8
Poor value (46.8). Costs more per task than its benchmark results predict. Plenty of data behind it.
48 ±9
Typical (48 ±9). Generates about as many tokens as models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
Not measured
GLM 4.6
🇨🇳
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 versionGLM 5.3We compared release dates and version names in our catalogue.
🇺🇸 US
Inference by Meganova
123.1
From 174 benchmarks. Plenty of data behind it.
96 ±22
Coding IQ 96.0, 95% interval 74–118. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.4725$1.995Not published by the provider
46.8
Poor value (46.8). Costs more per task than its benchmark results predict. Plenty of data behind it.
48 ±9
Typical (48 ±9). Generates about as many tokens as models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
Not measured
GLM 4.6
🇨🇳
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 versionGLM 5.3We compared release dates and version names in our catalogue.
z-ai/glm-4.6-original · original · 🇺🇸 US
Inference by Meganova
123.1
From 174 benchmarks. Plenty of data behind it.
96 ±22
Coding IQ 96.0, 95% interval 74–118. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.4725$1.995Not published by the provider
46.8
Poor value (46.8). Costs more per task than its benchmark results predict. Plenty of data behind it.
48 ±9
Typical (48 ±9). Generates about as many tokens as models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
Not measured
GLM 4.6
🇨🇳
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 versionGLM 5.3We compared release dates and version names in our catalogue.
z-ai/glm-4.6:thinking · thinking · 🇺🇸 US
Inference by Meganova
123.1
From 174 benchmarks. Plenty of data behind it.
96 ±22
Coding IQ 96.0, 95% interval 74–118. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.4725$1.995Not published by the provider
46.8
Poor value (46.8). Costs more per task than its benchmark results predict. Plenty of data behind it.
48 ±9
Typical (48 ±9). Generates about as many tokens as models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
Not measured
GLM 4.6
🇨🇳
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 versionGLM 5.3We compared release dates and version names in our catalogue.
bf16 · 🇺🇸 US
Inference by Novita AI
123.1
From 174 benchmarks. Plenty of data behind it.
96 ±22
Coding IQ 96.0, 95% interval 74–118. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.5198cache read $0.104$2.079Not published by the provider
46.8
Poor value (46.8). Costs more per task than its benchmark results predict. Plenty of data behind it.
48 ±9
Typical (48 ±9). Generates about as many tokens as models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
Not measured
GLM 4.6
🇨🇳
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 versionGLM 5.3We compared release dates and version names in our catalogue.
z-ai/glm-4.6-original · original · bf16 · 🇺🇸 US
Inference by Novita AI
123.1
From 174 benchmarks. Plenty of data behind it.
96 ±22
Coding IQ 96.0, 95% interval 74–118. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.5198cache read $0.104$2.079Not published by the provider
46.8
Poor value (46.8). Costs more per task than its benchmark results predict. Plenty of data behind it.
48 ±9
Typical (48 ±9). Generates about as many tokens as models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
Not measured
GLM 4.6
🇨🇳
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 versionGLM 5.3We compared release dates and version names in our catalogue.
z-ai/glm-4.6:thinking · thinking · bf16 · 🇺🇸 US
Inference by Novita AI
123.1
From 174 benchmarks. Plenty of data behind it.
96 ±22
Coding IQ 96.0, 95% interval 74–118. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.5198cache read $0.104$2.079Not published by the provider
46.8
Poor value (46.8). Costs more per task than its benchmark results predict. Plenty of data behind it.
48 ±9
Typical (48 ±9). Generates about as many tokens as models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
Not measured
GLM 4.6
🇨🇳
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 versionGLM 5.3We compared release dates and version names in our catalogue.
fp4 · 🇸🇬 SG
Inference by Z.ai
123.1
From 174 benchmarks. Plenty of data behind it.
96 ±22
Coding IQ 96.0, 95% interval 74–118. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.63cache read $0.1155$2.31Not published by the provider
46.8
Poor value (46.8). Costs more per task than its benchmark results predict. Plenty of data behind it.
48 ±9
Typical (48 ±9). Generates about as many tokens as models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
Not measured
GLM 4.6
🇨🇳
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 versionGLM 5.3We compared release dates and version names in our catalogue.
z-ai/glm-4.6-original · original · fp4 · 🇸🇬 SG
Inference by Z.ai
123.1
From 174 benchmarks. Plenty of data behind it.
96 ±22
Coding IQ 96.0, 95% interval 74–118. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.63cache read $0.1155$2.31Not published by the provider
46.8
Poor value (46.8). Costs more per task than its benchmark results predict. Plenty of data behind it.
48 ±9
Typical (48 ±9). Generates about as many tokens as models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
Not measured
GLM 4.6
🇨🇳
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 versionGLM 5.3We compared release dates and version names in our catalogue.
z-ai/glm-4.6:thinking · thinking · fp4 · 🇸🇬 SG
Inference by Z.ai
123.1
From 174 benchmarks. Plenty of data behind it.
96 ±22
Coding IQ 96.0, 95% interval 74–118. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.63cache read $0.1155$2.31Not published by the provider
46.8
Poor value (46.8). Costs more per task than its benchmark results predict. Plenty of data behind it.
48 ±9
Typical (48 ±9). Generates about as many tokens as models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
Not measured
123.0
From 7 benchmarks. A fair amount of data behind it.
≈ 106 ±10
Coding IQ 105.8, 95% interval 96–116. 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.
$0.25cache read $0.125$2.00262K
56.0
Good value (56.0). Cheaper per task than its benchmark results predict. Little data behind it, so treat it as a rough estimate.
Not measuredNot measured
MiniMax M2.5
🇨🇳
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.
122.8
From 192 benchmarks. Plenty of data behind it.
100 ±6
Coding IQ 100.4, 95% interval 94–106. Plenty 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$1.20205K
48.7
Fair value (48.7). Costs per task about what its benchmark results predict. Plenty of data behind it.
64 ±9
Chatty (64 ±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
Claude Haiku 4.5
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Newer version
Newer versionClaude Haiku 5.5We compared release dates and version names in our catalogue.
122.6
From 470 benchmarks. Plenty of data behind it.
95 ±6
Coding IQ 95.1, 95% interval 90–101. 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 $1.25200K
35.0
Poor value (35.0). Costs more per task than its benchmark results predict. Plenty of data behind it.
59 ±10
Chatty (59 ±10). Generates more tokens than models with similar coding results. Measured at default effort. A fair amount of data behind it.
142.7
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Claude Haiku 4.5
🇺🇸
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Newer version
Newer versionClaude Haiku 5.5We compared release dates and version names in our catalogue.
anthropic/claude-haiku-4.5:thinking · thinking
122.6
From 470 benchmarks. Plenty of data behind it.
95 ±6
Coding IQ 95.1, 95% interval 90–101. 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 $1.25200K
35.0
Poor value (35.0). Costs more per task than its benchmark results predict. Plenty of data behind it.
59 ±10
Chatty (59 ±10). Generates more tokens than models with similar coding results. Measured at default effort. A fair amount of data behind it.
142.7
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
MiniMax M2.1
🇨🇳
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.
122.3
From 97 benchmarks. Plenty of data behind it.
97 ±6
Coding IQ 96.5, 95% interval 91–102. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.33cache read $0.165$1.32200K
45.1
Poor value (45.1). Costs more per task than its benchmark results predict. Plenty of data behind it.
59 ±13
Chatty (59 ±13). 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
Qwen3.5 Omni Plus
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
122.2
From 62 benchmarks. Plenty of data behind it.
Not enough data yet——984KNot enough data yetNot measuredNot measured
DeepSeek R1
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
deepseek-ai/DeepSeek-R1-0528 · 0528
121.9
From 412 benchmarks. Plenty of data behind it.
93 ±14
Coding IQ 92.9, 95% interval 79–107. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.50cache read $0.35$2.15164K
36.1
Poor value (36.1). Costs more per task than its benchmark results predict. Plenty of data behind it.
60 ±11
Chatty (60 ±11). Generates more tokens than models with similar coding results. Measured at default effort. A fair amount of data behind it.
140.7
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
DeepSeek R1
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
deepseek-ai/DeepSeek-R1-0528 · 0528 · fp4 · 🇺🇸 US
Inference by DeepInfra
121.9
From 412 benchmarks. Plenty of data behind it.
93 ±14
Coding IQ 92.9, 95% interval 79–107. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.525cache read $0.3675$2.2575Not published by the provider
36.1
Poor value (36.1). Costs more per task than its benchmark results predict. Plenty of data behind it.
60 ±11
Chatty (60 ±11). Generates more tokens than models with similar coding results. Measured at default effort. A fair amount of data behind it.
140.7
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
DeepSeek R1
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
deepseek-ai/DeepSeek-R1-0528 · 0528 · 🇺🇸 US
Inference by Meganova
121.9
From 412 benchmarks. Plenty of data behind it.
93 ±14
Coding IQ 92.9, 95% interval 79–107. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.525cache read $0.42$2.2575Not published by the provider
36.1
Poor value (36.1). Costs more per task than its benchmark results predict. Plenty of data behind it.
60 ±11
Chatty (60 ±11). Generates more tokens than models with similar coding results. Measured at default effort. A fair amount of data behind it.
140.7
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
DeepSeek R1
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
deepseek-ai/DeepSeek-R1-0528 · 0528 · fp8 · 🇸🇬 SG
Inference by SiliconFlow
121.9
From 412 benchmarks. Plenty of data behind it.
93 ±14
Coding IQ 92.9, 95% interval 79–107. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.525$2.289Not published by the provider
36.1
Poor value (36.1). Costs more per task than its benchmark results predict. Plenty of data behind it.
60 ±11
Chatty (60 ±11). Generates more tokens than models with similar coding results. Measured at default effort. A fair amount of data behind it.
140.7
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
DeepSeek R1
🇨🇳
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
deepseek-ai/DeepSeek-R1-0528 · 0528 · 🇨🇳 CN
Inference by Streamlake
121.9
From 412 benchmarks. Plenty of data behind it.
93 ±14
Coding IQ 92.9, 95% interval 79–107. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.5996$2.4003Not published by the provider
36.1
Poor value (36.1). Costs more per task than its benchmark results predict. Plenty of data behind it.
60 ±11
Chatty (60 ±11). Generates more tokens than models with similar coding results. Measured at default effort. A fair amount of data behind it.
140.7
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GLM 4.5
🇨🇳
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 versionGLM 5.3We compared release dates and version names in our catalogue.
121.9
From 110 benchmarks. Plenty of data behind it.
95 ±7
Coding IQ 95.3, 95% interval 88–103. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.60cache read $0.11$2.20128K
41.4
Poor value (41.4). Costs more per task than its benchmark results predict. Plenty of data behind it.
53 ±9
Typical (53 ±9). Generates about as many tokens as models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
Not measured
GLM 4.5
🇨🇳
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 versionGLM 5.3We compared release dates and version names in our catalogue.
z-ai/GLM-4.5:thinking · thinking
121.9
From 110 benchmarks. Plenty of data behind it.
95 ±7
Coding IQ 95.3, 95% interval 88–103. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.60cache read $0.11$2.20128K
41.4
Poor value (41.4). Costs more per task than its benchmark results predict. Plenty of data behind it.
53 ±9
Typical (53 ±9). Generates about as many tokens as models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
Not measured

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

How this data works