Skip to content

Listings

All listings: OpenRouter

Every listing for the models on page 8 of 26 of OpenRouter, including each model's other versions and routes. Back to OpenRouter

Sold by OpenRouter

73 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 2.5 Flash
πŸ‡ΊπŸ‡Έ
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 the version names in our catalogue.
121.6
From 327 benchmarks. Plenty of data behind it.
86Β Β±16
Coding IQ 85.5, 95% interval 70–102. 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$2.501M
39.0
Poor value (39.0). Costs more per task than its benchmark results predict. Plenty of data behind it.
57 Β±9
Chatty (57 Β±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.
141.9
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Gemini 2.5 Flash
πŸ‡ΊπŸ‡Έ
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 the version names in our catalogue.
Inference by Google
121.6
From 327 benchmarks. Plenty of data behind it.
86Β Β±16
Coding IQ 85.5, 95% interval 70–102. 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$2.501M
39.0
Poor value (39.0). Costs more per task than its benchmark results predict. Plenty of data behind it.
57 Β±9
Chatty (57 Β±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.
141.9
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Gemini 2.5 Flash
πŸ‡ΊπŸ‡Έ
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 the version names in our catalogue.
flex
Inference by Google
121.6
From 327 benchmarks. Plenty of data behind it.
86Β Β±16
Coding IQ 85.5, 95% interval 70–102. 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 $0.015$1.251M
39.0
Poor value (39.0). Costs more per task than its benchmark results predict. Plenty of data behind it.
57 Β±9
Chatty (57 Β±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.
141.9
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Gemini 2.5 Flash
πŸ‡ΊπŸ‡Έ
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 the version names in our catalogue.
priority
Inference by Google
121.6
From 327 benchmarks. Plenty of data behind it.
86Β Β±16
Coding IQ 85.5, 95% interval 70–102. A fair amount of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.54cache read $0.054$4.501M
39.0
Poor value (39.0). Costs more per task than its benchmark results predict. Plenty of data behind it.
57 Β±9
Chatty (57 Β±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.
141.9
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
DeepSeek V3.1 Terminus
πŸ‡¨πŸ‡³
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
121.4
From 57 benchmarks. Plenty of data behind it.
β‰ˆ 99Β Β±13
Coding IQ 99.4, 95% interval 86–112. 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.
$0.27$1.00164K
57.6
Good value (57.6). Cheaper per task than its benchmark results predict. A fair amount of data behind it.
50 Β±9
Typical (50 Β±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
DeepSeek V3.1 Terminus
πŸ‡¨πŸ‡³
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
fp8
Inference by SiliconFlow
121.4
From 57 benchmarks. Plenty of data behind it.
β‰ˆ 99Β Β±13
Coding IQ 99.4, 95% interval 86–112. 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.
$0.27$1.00164K
57.6
Good value (57.6). Cheaper per task than its benchmark results predict. A fair amount of data behind it.
50 Β±9
Typical (50 Β±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
Gemini 2.5 Flash Preview 05-20 (no think)
πŸ‡ΊπŸ‡Έ
Maker headquartered in US, according to the company's own website. The citation is on the model page.
gemini-2.5-flash-preview-05-20:thinking Β· thinking
121.2
From 60 benchmarks. Plenty of data behind it.
Not enough data yet$0.15cache read $0.0375$3.50cache write $0.2333Not published by the provider
38.2
Poor value (38.2). Costs more per task than its benchmark results predict. A fair amount of data behind it.
Not measuredNot measured
Qwen3.5 35B A3B
πŸ‡¨πŸ‡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.
Newer version
Newer versionQwen3.6 35B A3BWe compared release dates and version names in our catalogue.
121.1
From 311 benchmarks. Plenty of data behind it.
95Β Β±8
Coding IQ 95.0, 95% interval 87–103. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.14cache read $0.05$1.00262K
40.3
Poor value (40.3). Costs more per task than its benchmark results predict. Plenty of data behind it.
59 Β±9
Chatty (59 Β±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.
144.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Qwen3.5 35B A3B
πŸ‡¨πŸ‡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.
Newer version
Newer versionQwen3.6 35B A3BWe compared release dates and version names in our catalogue.
Inference by Alibaba Cloud
121.1
From 311 benchmarks. Plenty of data behind it.
95Β Β±8
Coding IQ 95.0, 95% interval 87–103. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.1625$1.30258K
40.3
Poor value (40.3). Costs more per task than its benchmark results predict. Plenty of data behind it.
59 Β±9
Chatty (59 Β±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.
144.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Qwen3.5 35B A3B
πŸ‡¨πŸ‡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.
Newer version
Newer versionQwen3.6 35B A3BWe compared release dates and version names in our catalogue.
fp4
Inference by Darkbloom
121.1
From 311 benchmarks. Plenty of data behind it.
95Β Β±8
Coding IQ 95.0, 95% interval 87–103. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.065cache read $0.0325$0.75262K
40.3
Poor value (40.3). Costs more per task than its benchmark results predict. Plenty of data behind it.
59 Β±9
Chatty (59 Β±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.
144.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Qwen3.5 35B A3B
πŸ‡¨πŸ‡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.
Newer version
Newer versionQwen3.6 35B A3BWe compared release dates and version names in our catalogue.
fp8
Inference by DeepInfra
121.1
From 311 benchmarks. Plenty of data behind it.
95Β Β±8
Coding IQ 95.0, 95% interval 87–103. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.14cache read $0.05$1.00262K
40.3
Poor value (40.3). Costs more per task than its benchmark results predict. Plenty of data behind it.
59 Β±9
Chatty (59 Β±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.
144.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Qwen3.5 35B A3B
πŸ‡¨πŸ‡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.
Newer version
Newer versionQwen3.6 35B A3BWe compared release dates and version names in our catalogue.
fp8/flex Β· fp8
Inference by DeepInfra
121.1
From 311 benchmarks. Plenty of data behind it.
95Β Β±8
Coding IQ 95.0, 95% interval 87–103. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.112cache read $0.04$0.80262K
40.3
Poor value (40.3). Costs more per task than its benchmark results predict. Plenty of data behind it.
59 Β±9
Chatty (59 Β±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.
144.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Qwen3.5 35B A3B
πŸ‡¨πŸ‡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.
Newer version
Newer versionQwen3.6 35B A3BWe compared release dates and version names in our catalogue.
fp8
Inference by Parasail
121.1
From 311 benchmarks. Plenty of data behind it.
95Β Β±8
Coding IQ 95.0, 95% interval 87–103. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.15cache read $0.05$1.00262K
40.3
Poor value (40.3). Costs more per task than its benchmark results predict. Plenty of data behind it.
59 Β±9
Chatty (59 Β±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.
144.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Qwen3.5 35B A3B
πŸ‡¨πŸ‡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.
Newer version
Newer versionQwen3.6 35B A3BWe compared release dates and version names in our catalogue.
fp8
Inference by SiliconFlow
121.1
From 311 benchmarks. Plenty of data behind it.
95Β Β±8
Coding IQ 95.0, 95% interval 87–103. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.24cache read $0.15$1.80262K
40.3
Poor value (40.3). Costs more per task than its benchmark results predict. Plenty of data behind it.
59 Β±9
Chatty (59 Β±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.
144.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Qwen3.5 35B A3B
πŸ‡¨πŸ‡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.
Newer version
Newer versionQwen3.6 35B A3BWe compared release dates and version names in our catalogue.
Inference by Venice
121.1
From 311 benchmarks. Plenty of data behind it.
95Β Β±8
Coding IQ 95.0, 95% interval 87–103. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.15cache read $0.05$1.00256K
40.3
Poor value (40.3). Costs more per task than its benchmark results predict. Plenty of data behind it.
59 Β±9
Chatty (59 Β±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.
144.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
ERNIE 4.5 VL 424B A47B
πŸ‡¨πŸ‡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.
β‰ˆ 121
From 3 benchmarks. Little data behind it, so treat it as a rough estimate.
Not enough data yet$0.42$1.25Not published by the provider
54.2
Good value (54.2). Cheaper per task than its benchmark results predict. Little data behind it, so treat it as a rough estimate.
Not measuredNot measured
Gemini 2.5 Flash Preview
πŸ‡ΊπŸ‡Έ
Maker headquartered in US, according to the company's own website. The citation is on the model page.
120.8
From 67 benchmarks. A fair amount of data behind it.
Not enough data yet$0.15cache read $0.0375$0.60cache write $0.2333Not published by the providerNot enough data yetNot measuredNot measured
Gemini 2.5 Flash Preview
πŸ‡ΊπŸ‡Έ
Maker headquartered in US, according to the company's own website. The citation is on the model page.
gemini-2.5-flash-preview:thinking Β· thinking
120.8
From 67 benchmarks. A fair amount of data behind it.
Not enough data yet$0.15cache read $0.0375$3.50cache write $0.2333Not published by the providerNot enough data yetNot measuredNot measured
GPT 4.1
πŸ‡ΊπŸ‡Έ
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.
120.7
From 486 benchmarks. Plenty of data behind it.
92Β Β±6
Coding IQ 92.4, 95% interval 87–98. 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$8.001M
42.8
Poor value (42.8). Costs more per task than its benchmark results predict. Plenty of data behind it.
33 Β±8
Terse (33 Β±8). Generates fewer tokens than models with similar coding results. Measured at default effort. A fair amount of data behind it.
136.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GPT 4.1
πŸ‡ΊπŸ‡Έ
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-4.1:batch Β· batch
120.7
From 486 benchmarks. Plenty of data behind it.
92Β Β±6
Coding IQ 92.4, 95% interval 87–98. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$1.00cache read $0.25$4.001M
42.8
Poor value (42.8). Costs more per task than its benchmark results predict. Plenty of data behind it.
33 Β±8
Terse (33 Β±8). Generates fewer tokens than models with similar coding results. Measured at default effort. A fair amount of data behind it.
136.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GPT 4.1
πŸ‡ΊπŸ‡Έ
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
120.7
From 486 benchmarks. Plenty of data behind it.
92Β Β±6
Coding IQ 92.4, 95% interval 87–98. 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$8.001M
42.8
Poor value (42.8). Costs more per task than its benchmark results predict. Plenty of data behind it.
33 Β±8
Terse (33 Β±8). Generates fewer tokens than models with similar coding results. Measured at default effort. A fair amount of data behind it.
136.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GPT 4.1
πŸ‡ΊπŸ‡Έ
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
120.7
From 486 benchmarks. Plenty of data behind it.
92Β Β±6
Coding IQ 92.4, 95% interval 87–98. 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$8.001M
42.8
Poor value (42.8). Costs more per task than its benchmark results predict. Plenty of data behind it.
33 Β±8
Terse (33 Β±8). Generates fewer tokens than models with similar coding results. Measured at default effort. A fair amount of data behind it.
136.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Qwen3 235B A22B Instruct 2507
πŸ‡¨πŸ‡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.
qwen/qwen3-235b-a22b-2507
120.7
From 258 benchmarks. Plenty of data behind it.
92Β Β±11
Coding IQ 92.4, 95% interval 82–103. A fair amount of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.14cache read $0.05$0.80131K
58.8
Good value (58.8). Cheaper per task than its benchmark results predict. Plenty of data behind it.
Not measured
138
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Qwen3 235B A22B Instruct 2507
πŸ‡¨πŸ‡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.
qwen/qwen3-235b-a22b-2507 Β· fp8/flex Β· fp8
Inference by DeepInfra
120.7
From 258 benchmarks. Plenty of data behind it.
92Β Β±11
Coding IQ 92.4, 95% interval 82–103. A fair amount of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.072$0.44262K
58.8
Good value (58.8). Cheaper per task than its benchmark results predict. Plenty of data behind it.
Not measured
138
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Qwen3 235B A22B Instruct 2507
πŸ‡¨πŸ‡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.
qwen/qwen3-235b-a22b-2507 Β· fp8
Inference by Parasail
120.7
From 258 benchmarks. Plenty of data behind it.
92Β Β±11
Coding IQ 92.4, 95% interval 82–103. A fair amount of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.14cache read $0.05$0.80131K
58.8
Good value (58.8). Cheaper per task than its benchmark results predict. Plenty of data behind it.
Not measured
138
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Qwen3 235B A22B Instruct 2507
πŸ‡¨πŸ‡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.
qwen/qwen3-235b-a22b-2507 Β· fp8
Inference by Venice
120.7
From 258 benchmarks. Plenty of data behind it.
92Β Β±11
Coding IQ 92.4, 95% interval 82–103. A fair amount of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.15$0.75128K
58.8
Good value (58.8). Cheaper per task than its benchmark results predict. Plenty of data behind it.
Not measured
138
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Grok 3 Mini
πŸ‡ΊπŸ‡Έ
Maker headquartered in US, according to the company's own website. The citation is on the model page.
120.4
From 190 benchmarks. Plenty of data behind it.
β‰ˆ 95Β Β±12
Coding IQ 94.5, 95% interval 82–107. 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.
$0.30cache read $0.075$0.50Not published by the provider
50.6
Fair value (50.6). Costs per task about what its benchmark results predict. Plenty of data behind it.
Not measured
139.9
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Grok 3 Mini
πŸ‡ΊπŸ‡Έ
Maker headquartered in US, according to the company's own website. The citation is on the model page.
x-ai/grok-3-mini-beta Β· beta
120.4
From 190 benchmarks. Plenty of data behind it.
β‰ˆ 95Β Β±12
Coding IQ 94.5, 95% interval 82–107. 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.
$0.30$0.50Not published by the provider
50.6
Fair value (50.6). Costs per task about what its benchmark results predict. Plenty of data behind it.
Not measured
139.9
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Qwen3 VL 235B
πŸ‡¨πŸ‡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.
qwen/qwen3-vl-235b-a22b-thinking Β· thinking
120.4
From 170 benchmarks. Plenty of data behind it.
β‰ˆ 94Β Β±18
Coding IQ 94.2, 95% interval 76–113. 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.
$0.26$2.60Not published by the provider
45.0
Poor value (45.0). Costs more per task than its benchmark results predict. A fair amount of data behind it.
Not measuredNot measured
Qwen3 VL 235B
πŸ‡¨πŸ‡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.
qwen/qwen3-vl-235b-a22b-instruct
120.4
From 170 benchmarks. Plenty of data behind it.
β‰ˆ 94Β Β±18
Coding IQ 94.2, 95% interval 76–113. 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.
$0.21cache read $0.10$1.90131K
45.0
Poor value (45.0). Costs more per task than its benchmark results predict. A fair amount of data behind it.
Not measuredNot measured
Qwen3 VL 235B
πŸ‡¨πŸ‡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.
qwen/qwen3-vl-235b-a22b-instruct Β· fp8
Inference by DeepInfra
120.4
From 170 benchmarks. Plenty of data behind it.
β‰ˆ 94Β Β±18
Coding IQ 94.2, 95% interval 76–113. 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.
$0.20cache read $0.11$0.88262K
45.0
Poor value (45.0). Costs more per task than its benchmark results predict. A fair amount of data behind it.
Not measuredNot measured
Qwen3 VL 235B
πŸ‡¨πŸ‡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.
qwen/qwen3-vl-235b-a22b-instruct Β· fp8
Inference by Parasail
120.4
From 170 benchmarks. Plenty of data behind it.
β‰ˆ 94Β Β±18
Coding IQ 94.2, 95% interval 76–113. 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.
$0.21cache read $0.10$1.90131K
45.0
Poor value (45.0). Costs more per task than its benchmark results predict. A fair amount of data behind it.
Not measuredNot measured
Qwen3 VL 235B
πŸ‡¨πŸ‡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.
qwen/qwen3-vl-235b-a22b-instruct Β· fp8
Inference by Venice
120.4
From 170 benchmarks. Plenty of data behind it.
β‰ˆ 94Β Β±18
Coding IQ 94.2, 95% interval 76–113. 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.
$0.21cache read $0.10$1.90128K
45.0
Poor value (45.0). Costs more per task than its benchmark results predict. A fair amount of data behind it.
Not measuredNot measured
GPT 5.4 Nano
πŸ‡ΊπŸ‡Έ
Maker headquartered in US, according to the company's own website. The citation is on the model page.
120.3
From 252 benchmarks. Plenty of data behind it.
104Β Β±6
Coding IQ 103.6, 95% interval 97–110. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.20cache read $0.02$1.25400K
49.5
Fair value (49.5). Costs per task about what its benchmark results predict. Plenty of data behind it.
47 Β±9
Typical (47 Β±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.
146.4
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GPT 5.4 Nano
πŸ‡ΊπŸ‡Έ
Maker headquartered in US, according to the company's own website. The citation is on the model page.
openai/gpt-5.4-nano:batch Β· batch
120.3
From 252 benchmarks. Plenty of data behind it.
104Β Β±6
Coding IQ 103.6, 95% interval 97–110. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.10cache read $0.01$0.625400K
49.5
Fair value (49.5). Costs per task about what its benchmark results predict. Plenty of data behind it.
47 Β±9
Typical (47 Β±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.
146.4
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GPT 5.4 Nano
πŸ‡ΊπŸ‡Έ
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Inference by Microsoft Azure
120.3
From 252 benchmarks. Plenty of data behind it.
104Β Β±6
Coding IQ 103.6, 95% interval 97–110. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.20cache read $0.02$1.25400K
49.5
Fair value (49.5). Costs per task about what its benchmark results predict. Plenty of data behind it.
47 Β±9
Typical (47 Β±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.
146.4
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GPT 5.4 Nano
πŸ‡ΊπŸ‡Έ
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Inference by OpenAI
120.3
From 252 benchmarks. Plenty of data behind it.
104Β Β±6
Coding IQ 103.6, 95% interval 97–110. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.20cache read $0.02$1.25400K
49.5
Fair value (49.5). Costs per task about what its benchmark results predict. Plenty of data behind it.
47 Β±9
Typical (47 Β±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.
146.4
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GPT 5.4 Nano
πŸ‡ΊπŸ‡Έ
Maker headquartered in US, according to the company's own website. The citation is on the model page.
flex
Inference by OpenAI
120.3
From 252 benchmarks. Plenty of data behind it.
104Β Β±6
Coding IQ 103.6, 95% interval 97–110. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.10cache read $0.01$0.625400K
49.5
Fair value (49.5). Costs per task about what its benchmark results predict. Plenty of data behind it.
47 Β±9
Typical (47 Β±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.
146.4
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Qwen3 Coder Plus
πŸ‡¨πŸ‡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.
120.2
From 11 benchmarks. A fair amount of data behind it.
Not enough data yet$0.65cache read $0.13$3.25cache write $0.8125Not published by the provider
46.0
Poor value (46.0). Costs more per task than its benchmark results predict. Little data behind it, so treat it as a rough estimate.
39 Β±9
Terse (39 Β±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.
Not measured
Qwen3 Next 80B A3B
πŸ‡¨πŸ‡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.
qwen/qwen3-next-80b-a3b-thinking Β· thinking
120.2
From 191 benchmarks. Plenty of data behind it.
90Β Β±12
Coding IQ 90.0, 95% interval 78–102. A fair amount of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.0975$0.78Not published by the provider
51.1
Fair value (51.1). Costs per task about what its benchmark results predict. Little data behind it, so treat it as a rough estimate.
Not measuredNot measured
Step 3.7 Flash
πŸ‡¨πŸ‡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.
120.2
From 41 benchmarks. Plenty of data behind it.
106Β Β±15
Coding IQ 105.6, 95% interval 90–121. A fair amount of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.20cache read $0.04$1.15256K
47.0
Poor value (47.0). Costs more per task than its benchmark results predict. Plenty of data behind it.
57 Β±9
Chatty (57 Β±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
Step 3.7 Flash
πŸ‡¨πŸ‡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
120.2
From 41 benchmarks. Plenty of data behind it.
106Β Β±15
Coding IQ 105.6, 95% interval 90–121. A fair amount of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.20cache read $0.04$1.15262K
47.0
Poor value (47.0). Costs more per task than its benchmark results predict. Plenty of data behind it.
57 Β±9
Chatty (57 Β±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
Step 3.7 Flash
πŸ‡¨πŸ‡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 StepFun
120.2
From 41 benchmarks. Plenty of data behind it.
106Β Β±15
Coding IQ 105.6, 95% interval 90–121. A fair amount of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.20cache read $0.04$1.15256K
47.0
Poor value (47.0). Costs more per task than its benchmark results predict. Plenty of data behind it.
57 Β±9
Chatty (57 Β±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
Gemini 3 Pro Image
πŸ‡ΊπŸ‡Έ
Maker headquartered in US, according to the company's own website. The citation is on the model page.
119.8
From 189 benchmarks. A fair amount of data behind it.
Not enough data yet$2.00cache read $0.20$12.0066KNot enough data yetNot measuredNot measured
Gemini 3 Pro Image
πŸ‡ΊπŸ‡Έ
Maker headquartered in US, according to the company's own website. The citation is on the model page.
global
Inference by Google
119.8
From 189 benchmarks. A fair amount of data behind it.
Not enough data yet$2.00cache read $0.20$12.00131KNot enough data yetNot measuredNot measured
Gemini 3 Pro Image
πŸ‡ΊπŸ‡Έ
Maker headquartered in US, according to the company's own website. The citation is on the model page.
google/gemini-3-pro-image-preview Β· global/flex
Inference by Google
119.8
From 189 benchmarks. A fair amount of data behind it.
Not enough data yet$1.00cache read $0.10$6.0066KNot enough data yetNot measuredNot measured
Ling 1t
πŸ‡¨πŸ‡³
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
119.8
From 14 benchmarks. A fair amount of data behind it.
Not enough data yet$0.40$2.00Not published by the provider
58.0
Good value (58.0). Cheaper per task than its benchmark results predict. Little data behind it, so treat it as a rough estimate.
Not measuredNot measured
MiniMax M2.7
πŸ‡¨πŸ‡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.
119.8
From 207 benchmarks. Plenty of data behind it.
98Β Β±12
Coding IQ 98.4, 95% interval 87–110. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.60cache read $0.054$1.80197K
50.7
Fair value (50.7). Costs per task about what its benchmark results predict. Plenty of data behind it.
61 Β±9
Chatty (61 Β±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
MiniMax M2.7
πŸ‡¨πŸ‡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 Groq
119.8
From 207 benchmarks. Plenty of data behind it.
98Β Β±12
Coding IQ 98.4, 95% interval 87–110. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.60$1.80197K
50.7
Fair value (50.7). Costs per task about what its benchmark results predict. Plenty of data behind it.
61 Β±9
Chatty (61 Β±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
Qwen3.6 35B A3B
πŸ‡¨πŸ‡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.
119.8
From 160 benchmarks. Plenty of data behind it.
95Β Β±14
Coding IQ 94.8, 95% interval 81–109. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.10cache read $0.025$0.90262K
38.5
Poor value (38.5). Costs more per task than its benchmark results predict. Plenty of data behind it.
Not measured
144.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Qwen3.6 35B A3B
πŸ‡¨πŸ‡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 Akash
119.8
From 160 benchmarks. Plenty of data behind it.
95Β Β±14
Coding IQ 94.8, 95% interval 81–109. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.10cache read $0.05$0.90262K
38.5
Poor value (38.5). Costs more per task than its benchmark results predict. Plenty of data behind it.
Not measured
144.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Qwen3.6 35B A3B
πŸ‡¨πŸ‡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 Darkbloom
119.8
From 160 benchmarks. Plenty of data behind it.
95Β Β±14
Coding IQ 94.8, 95% interval 81–109. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.05cache read $0.025$0.70262K
38.5
Poor value (38.5). Costs more per task than its benchmark results predict. Plenty of data behind it.
Not measured
144.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Qwen3.6 35B A3B
πŸ‡¨πŸ‡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 DeepInfra
119.8
From 160 benchmarks. Plenty of data behind it.
95Β Β±14
Coding IQ 94.8, 95% interval 81–109. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.10$0.95262K
38.5
Poor value (38.5). Costs more per task than its benchmark results predict. Plenty of data behind it.
Not measured
144.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Qwen3.6 35B A3B
πŸ‡¨πŸ‡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/flex Β· fp8
Inference by DeepInfra
119.8
From 160 benchmarks. Plenty of data behind it.
95Β Β±14
Coding IQ 94.8, 95% interval 81–109. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.08$0.76262K
38.5
Poor value (38.5). Costs more per task than its benchmark results predict. Plenty of data behind it.
Not measured
144.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Qwen3.6 35B A3B
πŸ‡¨πŸ‡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/priority Β· fp8
Inference by DeepInfra
119.8
From 160 benchmarks. Plenty of data behind it.
95Β Β±14
Coding IQ 94.8, 95% interval 81–109. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.15$1.425262K
38.5
Poor value (38.5). Costs more per task than its benchmark results predict. Plenty of data behind it.
Not measured
144.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Qwen3.6 35B A3B
πŸ‡¨πŸ‡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 Parasail
119.8
From 160 benchmarks. Plenty of data behind it.
95Β Β±14
Coding IQ 94.8, 95% interval 81–109. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.10cache read $0.025$0.90262K
38.5
Poor value (38.5). Costs more per task than its benchmark results predict. Plenty of data behind it.
Not measured
144.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Qwen3.6 35B A3B
πŸ‡¨πŸ‡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 Primeintellect
119.8
From 160 benchmarks. Plenty of data behind it.
95Β Β±14
Coding IQ 94.8, 95% interval 81–109. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.19cache read $0.17$1.70262K
38.5
Poor value (38.5). Costs more per task than its benchmark results predict. Plenty of data behind it.
Not measured
144.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Qwen3.6 35B A3B
πŸ‡¨πŸ‡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
119.8
From 160 benchmarks. Plenty of data behind it.
95Β Β±14
Coding IQ 94.8, 95% interval 81–109. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.24cache read $0.15$1.80262K
38.5
Poor value (38.5). Costs more per task than its benchmark results predict. Plenty of data behind it.
Not measured
144.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Qwen3.6 35B A3B
πŸ‡¨πŸ‡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 Venice
119.8
From 160 benchmarks. Plenty of data behind it.
95Β Β±14
Coding IQ 94.8, 95% interval 81–109. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.10$1.00256K
38.5
Poor value (38.5). Costs more per task than its benchmark results predict. Plenty of data behind it.
Not measured
144.1
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GPT 4 32k
πŸ‡ΊπŸ‡Έ
Maker headquartered in US, according to the company's own website. The citation is on the model page.
β‰ˆ 120
From 4 benchmarks. Little data behind it, so treat it as a rough estimate.
Not enough data yet$60.00$120.00Not published by the providerNot enough data yetNot measuredNot measured
GPT-4.5 Preview
πŸ‡ΊπŸ‡Έ
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Newer version
Newer versionGPT-5.5-previewWe compared the version names in our catalogue.
119.7
From 206 benchmarks. Plenty of data behind it.
88Β Β±11
Coding IQ 88.3, 95% interval 77–100. A fair amount of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$75.00cache read $37.50$150.00Not published by the provider
23.1
Poor value (23.1). Costs more per task than its benchmark results predict. A fair amount of data behind it.
Not measured
136.5
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
DeepSeek V3.1
πŸ‡¨πŸ‡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.
deepseek/deepseek-chat-v3.1:free Β· free
119.6
From 122 benchmarks. Plenty of data behind it.
94Β Β±8
Coding IQ 94.3, 95% interval 86–103. A fair amount of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.00$0.00Not published by the provider
44.0
Poor value (44.0). Costs more per task than its benchmark results predict. Plenty of data behind it.
Not measured
140.9
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
DeepSeek V3.1
πŸ‡¨πŸ‡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.
deepseek/deepseek-chat-v3.1
119.6
From 122 benchmarks. Plenty of data behind it.
94Β Β±8
Coding IQ 94.3, 95% interval 86–103. A fair amount of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.25cache read $0.13$0.95164K
44.0
Poor value (44.0). Costs more per task than its benchmark results predict. Plenty of data behind it.
Not measured
140.9
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
DeepSeek V3.1
πŸ‡¨πŸ‡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.
deepseek/deepseek-chat-v3.1 Β· fp8
Inference by Coreweave
119.6
From 122 benchmarks. Plenty of data behind it.
94Β Β±8
Coding IQ 94.3, 95% interval 86–103. A fair amount of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.50cache read $0.20$1.50161K
44.0
Poor value (44.0). Costs more per task than its benchmark results predict. Plenty of data behind it.
Not measured
140.9
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
DeepSeek V3.1
πŸ‡¨πŸ‡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.
deepseek/deepseek-chat-v3.1 Β· fp4
Inference by DeepInfra
119.6
From 122 benchmarks. Plenty of data behind it.
94Β Β±8
Coding IQ 94.3, 95% interval 86–103. A fair amount of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.25cache read $0.13$0.95164K
44.0
Poor value (44.0). Costs more per task than its benchmark results predict. Plenty of data behind it.
Not measured
140.9
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
DeepSeek V3.1
πŸ‡¨πŸ‡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.
deepseek/deepseek-chat-v3.1 Β· fp8
Inference by SiliconFlow
119.6
From 122 benchmarks. Plenty of data behind it.
94Β Β±8
Coding IQ 94.3, 95% interval 86–103. A fair amount of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.27$1.00164K
44.0
Poor value (44.0). Costs more per task than its benchmark results predict. Plenty of data behind it.
Not measured
140.9
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Mistral Medium 3.1
πŸ‡«πŸ‡·
Maker headquartered in FR, according to the company's own website. The citation is on the model page.
Newer version
Newer versionMistral Medium 3.5We compared the version names in our catalogue.
119.6
From 89 benchmarks. Plenty of data behind it.
Not enough data yet$0.44cache read $0.044$2.20131KNot enough data yetNot measuredNot measured
Mistral Medium 3.1
πŸ‡«πŸ‡·
Maker headquartered in FR, according to the company's own website. The citation is on the model page.
Newer version
Newer versionMistral Medium 3.5We compared the version names in our catalogue.
eu
Inference by Mistral AI
119.6
From 89 benchmarks. Plenty of data behind it.
Not enough data yet$0.44cache read $0.044$2.20131KNot enough data yetNot measuredNot measured
Devstral 2 123b Instruct 2512
πŸ‡«πŸ‡·
Maker headquartered in FR, 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.
mistralai/devstral-2512
119.5
From 25 benchmarks. Plenty of data behind it.
97Β Β±9
Coding IQ 97.1, 95% interval 88–106. A fair amount of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.40cache read $0.04$2.00262K
50.2
Fair value (50.2). Costs per task about what its benchmark results predict. A fair amount of data behind it.
Not measuredNot measured
Devstral 2 123b Instruct 2512
πŸ‡«πŸ‡·
Maker headquartered in FR, 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.
mistralai/devstral-2512 Β· eu
Inference by Mistral AI
119.5
From 25 benchmarks. Plenty of data behind it.
97Β Β±9
Coding IQ 97.1, 95% interval 88–106. A fair amount of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.40cache read $0.04$2.00262K
50.2
Fair value (50.2). Costs per task about what its benchmark results predict. A fair amount of data behind it.
Not measuredNot measured
Gemini 2.5 Flash Lite Preview 09-2025
πŸ‡ΊπŸ‡Έ
Maker headquartered in US, according to the company's own website. The citation is on the model page.
119.5
From 21 benchmarks. A fair amount of data behind it.
Not enough data yet$0.10$0.40Not published by the provider
65.8
Good value (65.8). Cheaper per task than its benchmark results predict. Plenty of data behind it.
Not measuredNot measured
Kimi K2 0905
πŸ‡¨πŸ‡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.
Newer version
Newer versionKimi K3.1We compared the version names in our catalogue.
119.4
From 22 benchmarks. Plenty of data behind it.
β‰ˆ 93Β Β±11
Coding IQ 92.5, 95% interval 82–103. 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.
$0.60$2.50262K
56.6
Good value (56.6). Cheaper per task than its benchmark results predict. A fair amount of data behind it.
41 Β±9
Terse (41 Β±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.
Not measured
Kimi K2 0905
πŸ‡¨πŸ‡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.
Newer version
Newer versionKimi K3.1We compared the version names in our catalogue.
fp8
Inference by Novita AI
119.4
From 22 benchmarks. Plenty of data behind it.
β‰ˆ 93Β Β±11
Coding IQ 92.5, 95% interval 82–103. 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.
$0.60$2.50262K
56.6
Good value (56.6). Cheaper per task than its benchmark results predict. A fair amount of data behind it.
41 Β±9
Terse (41 Β±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.
Not measured

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

How this data works