Skip to content

Listings

All listings: OpenRouter

Every listing for the models on page 12 of 26 of OpenRouter, including each model's other versions and routes. Back to OpenRouter

Sold by OpenRouter

58 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
Qwen Max
πŸ‡¨πŸ‡³
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
111.6
From 56 benchmarks. Plenty of data behind it.
β‰ˆ 82Β Β±14
Coding IQ 81.7, 95% interval 68–95. 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.
$1.60cache read $0.64$6.40Not published by the provider
42.6
Poor value (42.6). Costs more per task than its benchmark results predict. Little data behind it, so treat it as a rough estimate.
Not measuredNot measured
qwen3-vl-30b
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-30b-a3b-thinking Β· thinking
111.5
From 81 benchmarks. Plenty of data behind it.
Not enough data yet$0.13$1.56Not published by the provider
45.6
Poor value (45.6). Costs more per task than its benchmark results predict. A fair amount of data behind it.
Not measuredNot measured
qwen3-vl-30b
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-30b-a3b-instruct
111.5
From 81 benchmarks. Plenty of data behind it.
Not enough data yet$0.15$0.60262K
45.6
Poor value (45.6). Costs more per task than its benchmark results predict. A fair amount of data behind it.
Not measuredNot measured
qwen3-vl-30b
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-30b-a3b-instruct Β· fp8
Inference by DeepInfra
111.5
From 81 benchmarks. Plenty of data behind it.
Not enough data yet$0.15$0.60262K
45.6
Poor value (45.6). Costs more per task than its benchmark results predict. A fair amount of data behind it.
Not measuredNot measured
qwen3-vl-30b
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-30b-a3b-instruct Β· bf16
Inference by Novita AI
111.5
From 81 benchmarks. Plenty of data behind it.
Not enough data yet$0.20$0.70131K
45.6
Poor value (45.6). Costs more per task than its benchmark results predict. A fair amount of data behind it.
Not measuredNot measured
qwen3-vl-30b
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-30b-a3b-instruct Β· fp8
Inference by SiliconFlow
111.5
From 81 benchmarks. Plenty of data behind it.
Not enough data yet$0.29$1.00262K
45.6
Poor value (45.6). Costs more per task than its benchmark results predict. A fair amount of data behind it.
Not measuredNot measured
GPT OSS 20b
πŸ‡ΊπŸ‡Έ
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.
openai/gpt-oss-20b:free Β· free
111.3
From 251 benchmarks. Plenty of data behind it.
88Β Β±12
Coding IQ 88.2, 95% interval 76–101. Plenty 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
48.9
Fair value (48.9). Costs per task about what its benchmark results predict. Plenty of data behind it.
40 Β±9
Terse (40 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.
134.9
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GPT OSS 20b
πŸ‡ΊπŸ‡Έ
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.
111.3
From 251 benchmarks. Plenty of data behind it.
88Β Β±12
Coding IQ 88.2, 95% interval 76–101. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.017cache read $0.0085$0.085131K
48.9
Fair value (48.9). Costs per task about what its benchmark results predict. Plenty of data behind it.
40 Β±9
Terse (40 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.
134.9
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GPT OSS 20b
πŸ‡ΊπŸ‡Έ
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.
openai/gpt-oss-20b:batch Β· batch
111.3
From 251 benchmarks. Plenty of data behind it.
88Β Β±12
Coding IQ 88.2, 95% interval 76–101. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.024$0.112131K
48.9
Fair value (48.9). Costs per task about what its benchmark results predict. Plenty of data behind it.
40 Β±9
Terse (40 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.
134.9
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GPT OSS 20b
πŸ‡ΊπŸ‡Έ
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.
fp4
Inference by Akash
111.3
From 251 benchmarks. Plenty of data behind it.
88Β Β±12
Coding IQ 88.2, 95% interval 76–101. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.018cache read $0.009$0.09131K
48.9
Fair value (48.9). Costs per task about what its benchmark results predict. Plenty of data behind it.
40 Β±9
Terse (40 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.
134.9
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GPT OSS 20b
πŸ‡ΊπŸ‡Έ
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.
Inference by Amazon Web Services
111.3
From 251 benchmarks. Plenty of data behind it.
88Β Β±12
Coding IQ 88.2, 95% interval 76–101. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.07$0.15131K
48.9
Fair value (48.9). Costs per task about what its benchmark results predict. Plenty of data behind it.
40 Β±9
Terse (40 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.
134.9
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GPT OSS 20b
πŸ‡ΊπŸ‡Έ
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.
eu-west-1
Inference by Amazon Web Services
111.3
From 251 benchmarks. Plenty of data behind it.
88Β Β±12
Coding IQ 88.2, 95% interval 76–101. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.07$0.15131K
48.9
Fair value (48.9). Costs per task about what its benchmark results predict. Plenty of data behind it.
40 Β±9
Terse (40 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.
134.9
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GPT OSS 20b
πŸ‡ΊπŸ‡Έ
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.
fp4
Inference by Coreweave
111.3
From 251 benchmarks. Plenty of data behind it.
88Β Β±12
Coding IQ 88.2, 95% interval 76–101. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.03$0.13131K
48.9
Fair value (48.9). Costs per task about what its benchmark results predict. Plenty of data behind it.
40 Β±9
Terse (40 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.
134.9
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GPT OSS 20b
πŸ‡ΊπŸ‡Έ
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.
fp8
Inference by Darkbloom
111.3
From 251 benchmarks. Plenty of data behind it.
88Β Β±12
Coding IQ 88.2, 95% interval 76–101. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.017cache read $0.0085$0.085131K
48.9
Fair value (48.9). Costs per task about what its benchmark results predict. Plenty of data behind it.
40 Β±9
Terse (40 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.
134.9
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GPT OSS 20b
πŸ‡ΊπŸ‡Έ
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.
openai/gpt-oss-20b:batch Β· batch+bf16 Β· bf16
Inference by DeepInfra
111.3
From 251 benchmarks. Plenty of data behind it.
88Β Β±12
Coding IQ 88.2, 95% interval 76–101. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.024$0.112131K
48.9
Fair value (48.9). Costs per task about what its benchmark results predict. Plenty of data behind it.
40 Β±9
Terse (40 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.
134.9
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GPT OSS 20b
πŸ‡ΊπŸ‡Έ
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.
bf16
Inference by DeepInfra
111.3
From 251 benchmarks. Plenty of data behind it.
88Β Β±12
Coding IQ 88.2, 95% interval 76–101. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.03cache read $0.015$0.14131K
48.9
Fair value (48.9). Costs per task about what its benchmark results predict. Plenty of data behind it.
40 Β±9
Terse (40 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.
134.9
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GPT OSS 20b
πŸ‡ΊπŸ‡Έ
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.
bf16/flex Β· bf16
Inference by DeepInfra
111.3
From 251 benchmarks. Plenty of data behind it.
88Β Β±12
Coding IQ 88.2, 95% interval 76–101. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.024cache read $0.012$0.112131K
48.9
Fair value (48.9). Costs per task about what its benchmark results predict. Plenty of data behind it.
40 Β±9
Terse (40 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.
134.9
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GPT OSS 20b
πŸ‡ΊπŸ‡Έ
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.
Inference by Groq
111.3
From 251 benchmarks. Plenty of data behind it.
88Β Β±12
Coding IQ 88.2, 95% interval 76–101. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.075cache read $0.0375$0.30131K
48.9
Fair value (48.9). Costs per task about what its benchmark results predict. Plenty of data behind it.
40 Β±9
Terse (40 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.
134.9
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GPT OSS 20b
πŸ‡ΊπŸ‡Έ
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.
fp4
Inference by Parasail
111.3
From 251 benchmarks. Plenty of data behind it.
88Β Β±12
Coding IQ 88.2, 95% interval 76–101. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.03cache read $0.02$0.15131K
48.9
Fair value (48.9). Costs per task about what its benchmark results predict. Plenty of data behind it.
40 Β±9
Terse (40 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.
134.9
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GPT OSS 20b
πŸ‡ΊπŸ‡Έ
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.
fp4
Inference by Primeintellect
111.3
From 251 benchmarks. Plenty of data behind it.
88Β Β±12
Coding IQ 88.2, 95% interval 76–101. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.20cache read $0.18$0.30131K
48.9
Fair value (48.9). Costs per task about what its benchmark results predict. Plenty of data behind it.
40 Β±9
Terse (40 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.
134.9
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GPT OSS 20b
πŸ‡ΊπŸ‡Έ
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.
fp8
Inference by SiliconFlow
111.3
From 251 benchmarks. Plenty of data behind it.
88Β Β±12
Coding IQ 88.2, 95% interval 76–101. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.04$0.18131K
48.9
Fair value (48.9). Costs per task about what its benchmark results predict. Plenty of data behind it.
40 Β±9
Terse (40 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.
134.9
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Qwen 2.5 VL 32B Instruct
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 VL 32B InstructWe compared release dates and version names in our catalogue.
111.3
From 25 benchmarks. A fair amount of data behind it.
Not enough data yetβ€”β€”Not published by the provider
47.9
Fair value (47.9). Costs per task about what its benchmark results predict. Little data behind it, so treat it as a rough estimate.
Not measuredNot measured
Qwen 2.5 VL 32B Instruct
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 VL 32B InstructWe compared release dates and version names in our catalogue.
qwen/qwen2.5-vl-32b-instruct:free Β· free
111.3
From 25 benchmarks. A fair amount of data behind it.
Not enough data yet$0.00$0.00Not published by the provider
47.9
Fair value (47.9). Costs per task about what its benchmark results predict. Little data behind it, so treat it as a rough estimate.
Not measuredNot measured
Claude 3 Opus
πŸ‡ΊπŸ‡Έ
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Newer version
Newer versionClaude Opus 5.5We compared release dates and version names in our catalogue.
111.1
From 307 benchmarks. Plenty of data behind it.
80Β Β±10
Coding IQ 79.6, 95% interval 70–89. A fair amount of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$15.00cache read $1.50$75.00cache write $18.75Not published by the provider
38.0
Poor value (38.0). Costs more per task than its benchmark results predict. Little data behind it, so treat it as a rough estimate.
Not measured
125.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Granite 4.1 8B
πŸ‡ΊπŸ‡Έ
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Newer version
Newer versionGranite 4.2 8BWe compared the version names in our catalogue.
β‰ˆ 111
From 36 benchmarks. Little data behind it, so treat it as a rough estimate.
Not enough data yet$0.05$0.10Not published by the provider
46.9
Poor value (46.9). Costs more per task than its benchmark results predict. Little data behind it, so treat it as a rough estimate.
Not measuredNot measured
Granite 4.2 8B
πŸ‡ΊπŸ‡Έ
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.
110.9
From 57 benchmarks. Plenty of data behind it.
90Β Β±17
Coding IQ 90.2, 95% interval 73–108. A fair amount of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.06cache read $0.015$0.15131KNot enough data yetNot measuredNot measured
Granite 4.2 8B
πŸ‡ΊπŸ‡Έ
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.
bf16
Inference by Coreweave
110.9
From 57 benchmarks. Plenty of data behind it.
90Β Β±17
Coding IQ 90.2, 95% interval 73–108. A fair amount of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.06cache read $0.015$0.15131KNot enough data yetNot measuredNot measured
Granite 4.2 8B
πŸ‡ΊπŸ‡Έ
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.
bf16
Inference by DeepInfra
110.9
From 57 benchmarks. Plenty of data behind it.
90Β Β±17
Coding IQ 90.2, 95% interval 73–108. A fair amount of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.06cache read $0.015$0.25131KNot enough data yetNot measuredNot measured
Nemotron Nano 9b V2
πŸ‡ΊπŸ‡Έ
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.
110.6
From 19 benchmarks. Plenty of data behind it.
Not enough data yet$0.04$0.16Not published by the provider
52.1
Fair value (52.1). Costs per task about what its benchmark results predict. Little data behind it, so treat it as a rough estimate.
Not measuredNot measured
Nemotron Nano 9b V2
πŸ‡ΊπŸ‡Έ
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-nano-9b-v2:free Β· free
110.6
From 19 benchmarks. Plenty of data behind it.
Not enough data yet$0.00$0.00Not published by the provider
52.1
Fair value (52.1). Costs per task about what its benchmark results predict. Little data behind it, so treat it as a rough estimate.
Not measuredNot measured
Hunyuan A13B Instruct
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 111
From 9 benchmarks. Little data behind it, so treat it as a rough estimate.
Not enough data yetβ€”β€”Not published by the provider
47.0
Poor value (47.0). Costs more per task than its benchmark results predict. Little data behind it, so treat it as a rough estimate.
Not measuredNot measured
Hunyuan A13B Instruct
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.
tencent/hunyuan-a13b-instruct:free Β· free
β‰ˆ 111
From 9 benchmarks. Little data behind it, so treat it as a rough estimate.
Not enough data yet$0.00$0.00Not published by the provider
47.0
Poor value (47.0). Costs more per task than its benchmark results predict. Little data behind it, so treat it as a rough estimate.
Not measuredNot measured
Claude Haiku 3.5
πŸ‡ΊπŸ‡Έ
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 the version names in our catalogue.
anthropic/claude-3.5-haiku
110.1
From 238 benchmarks. Plenty of data behind it.
79Β Β±8
Coding IQ 79.4, 95% interval 72–87. A fair amount of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.80$4.00Not published by the provider
19.4
Poor value (19.4). Costs more per task than its benchmark results predict. Plenty of data behind it.
Not measured
124.8
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Claude Haiku 3.5
πŸ‡ΊπŸ‡Έ
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 the version names in our catalogue.
anthropic/claude-3.5-haiku-20241022 Β· 20241022
110.1
From 238 benchmarks. Plenty of data behind it.
79Β Β±8
Coding IQ 79.4, 95% interval 72–87. A fair amount of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.80$4.00Not published by the provider
19.4
Poor value (19.4). Costs more per task than its benchmark results predict. Plenty of data behind it.
Not measured
124.8
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
DeepSeek R1 0528 Qwen3 8B
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 110
From 9 benchmarks. Little data behind it, so treat it as a rough estimate.
Not enough data yet$0.05$0.10Not published by the provider
43.4
Poor value (43.4). Costs more per task than its benchmark results predict. Little data behind it, so treat it as a rough estimate.
Not measuredNot measured
DeepSeek R1 0528 Qwen3 8B
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-r1-0528-qwen3-8b:free Β· free
β‰ˆ 110
From 9 benchmarks. Little data behind it, so treat it as a rough estimate.
Not enough data yet$0.00$0.00Not published by the provider
43.4
Poor value (43.4). Costs more per task than its benchmark results predict. Little data behind it, so treat it as a rough estimate.
Not measuredNot measured
β‰ˆ 110
From 21 benchmarks. Little data behind it, so treat it as a rough estimate.
Not enough data yet$0.25$1.00Not published by the provider
51.0
Fair value (51.0). Costs per task about what its benchmark results predict. Little data behind it, so treat it as a rough estimate.
Not measuredNot measured
109.8
From 29 benchmarks. Plenty of data behind it.
Not enough data yet$0.07$0.26Not published by the provider
53.1
Good value (53.1). Cheaper per task than its benchmark results predict. Little data behind it, so treat it as a rough estimate.
Not measuredNot measured
Llama 3.1 Nemotron 70B Instruct
πŸ‡ΊπŸ‡Έ
Maker headquartered in US, according to the company's own website. The citation is on the model page.
109.8
From 42 benchmarks. Plenty of data behind it.
Not enough data yet$0.12$0.30Not published by the provider
57.1
Good value (57.1). Cheaper per task than its benchmark results predict. Little data behind it, so treat it as a rough estimate.
Not measuredNot measured
MiniMax 01
πŸ‡¨πŸ‡³
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
β‰ˆ 110
From 8 benchmarks. Little data behind it, so treat it as a rough estimate.
Not enough data yet$0.20$1.10Not 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
Llama 3.1 405B
πŸ‡ΊπŸ‡Έ
Maker headquartered in US, according to the company's own website. The citation is on the model page.
109.7
From 26 benchmarks. Plenty of data behind it.
Not enough data yet$2.00$2.00Not published by the providerNot enough data yetNot measured
127.9
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Nemotron 3 Nano 30B A3B
πŸ‡ΊπŸ‡Έ
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.
109.7
From 137 benchmarks. Plenty of data behind it.
91Β Β±14
Coding IQ 91.4, 95% interval 77–106. A fair amount of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.06cache read $0.012$0.24262K
36.7
Poor value (36.7). Costs more per task than its benchmark results predict. Little data behind it, so treat it as a rough estimate.
Not measuredNot measured
Nemotron 3 Nano 30B A3B
πŸ‡ΊπŸ‡Έ
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.
fp8
Inference by Nebius AI Studio
109.7
From 137 benchmarks. Plenty of data behind it.
91Β Β±14
Coding IQ 91.4, 95% interval 77–106. A fair amount of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.06cache read $0.012$0.24262K
36.7
Poor value (36.7). Costs more per task than its benchmark results predict. Little data behind it, so treat it as a rough estimate.
Not measuredNot measured
Gemini 1.5 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.
gemini-flash-1.5
109.6
From 160 benchmarks. Plenty of data behind it.
β‰ˆ 77Β Β±12
Coding IQ 76.5, 95% interval 65–88. 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.
$0.075cache read $0.0188$0.30cache write $0.1583Not published by the providerNot enough data yetNot measured
127
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Phi 4 Reasoning Plus
πŸ‡ΊπŸ‡Έ
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.
109.5
From 16 benchmarks. A fair amount of data behind it.
Not enough data yet$0.07$0.35Not published by the provider
42.3
Poor value (42.3). Costs more per task than its benchmark results predict. Little data behind it, so treat it as a rough estimate.
Not measuredNot measured
Magistral Small 2509
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.
magistral-small-2506 Β· 2506
109.2
From 42 benchmarks. Plenty of data behind it.
Not enough data yet$0.50$1.50Not published by the provider
27.6
Poor value (27.6). Costs more per task than its benchmark results predict. Plenty of data behind it.
Not measuredNot measured
Gemini 3.1 Flash-Lite Image
πŸ‡ΊπŸ‡Έ
Maker headquartered in US, according to the company's own website. The citation is on the model page.
β‰ˆ 109
From 14 benchmarks. Little data behind it, so treat it as a rough estimate.
Not enough data yet$0.25$1.5066K
37.6
Poor value (37.6). Costs more per task than its benchmark results predict. Little data behind it, so treat it as a rough estimate.
Not measuredNot measured
Gemini 3.1 Flash-Lite Image
πŸ‡ΊπŸ‡Έ
Maker headquartered in US, according to the company's own website. The citation is on the model page.
Inference by Google
β‰ˆ 109
From 14 benchmarks. Little data behind it, so treat it as a rough estimate.
Not enough data yet$0.25$1.5066K
37.6
Poor value (37.6). Costs more per task than its benchmark results predict. Little data behind it, so treat it as a rough estimate.
Not measuredNot measured
Gemini 3.1 Flash-Lite 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
β‰ˆ 109
From 14 benchmarks. Little data behind it, so treat it as a rough estimate.
Not enough data yet$0.25$1.5066K
37.6
Poor value (37.6). Costs more per task than its benchmark results predict. Little data behind it, so treat it as a rough estimate.
Not measuredNot measured
Qwen Turbo
πŸ‡¨πŸ‡³
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
109.0
From 5 benchmarks. A fair amount of data behind it.
Not enough data yet$0.05cache read $0.02$0.20Not published by the provider
61.6
Good value (61.6). Cheaper per task than its benchmark results predict. Little data behind it, so treat it as a rough estimate.
Not measuredNot measured
Qwen: Qwen3 VL 8B Thinking
πŸ‡¨πŸ‡³
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
qwen/qwen3-vl-8b-thinking Β· thinking
109.0
From 103 benchmarks. Plenty of data behind it.
Not enough data yet$0.117$1.365Not published by the provider
41.1
Poor value (41.1). Costs more per task than its benchmark results predict. Little data behind it, so treat it as a rough estimate.
Not measuredNot measured
Qwen2.5 72B Instruct
πŸ‡¨πŸ‡³
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
qwen/qwen-2.5-72b-instruct:free Β· free
108.9
From 216 benchmarks. Plenty of data behind it.
80Β Β±22
Coding IQ 79.9, 95% interval 58–102. A fair amount of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.00$0.00Not published by the provider
35.0
Poor value (35.0). Costs more per task than its benchmark results predict. A fair amount of data behind it.
Not measuredNot measured
Qwen2.5 72B Instruct
πŸ‡¨πŸ‡³
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
108.9
From 216 benchmarks. Plenty of data behind it.
80Β Β±22
Coding IQ 79.9, 95% interval 58–102. A fair amount of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.36$0.4033K
35.0
Poor value (35.0). Costs more per task than its benchmark results predict. A fair amount of data behind it.
Not measuredNot measured
Qwen2.5 72B Instruct
πŸ‡¨πŸ‡³
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
fp8
Inference by DeepInfra
108.9
From 216 benchmarks. Plenty of data behind it.
80Β Β±22
Coding IQ 79.9, 95% interval 58–102. A fair amount of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.36$0.4033K
35.0
Poor value (35.0). Costs more per task than its benchmark results predict. A fair amount of data behind it.
Not measuredNot measured
Qwen2.5 72B Instruct
πŸ‡¨πŸ‡³
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
fp8/flex Β· fp8
Inference by DeepInfra
108.9
From 216 benchmarks. Plenty of data behind it.
80Β Β±22
Coding IQ 79.9, 95% interval 58–102. A fair amount of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.288$0.3233K
35.0
Poor value (35.0). Costs more per task than its benchmark results predict. A fair amount of data behind it.
Not measuredNot measured
Qwen2.5 VL 72B Instruct
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/qwen2.5-vl-72b-instruct:free Β· free
108.8
From 97 benchmarks. Plenty of data behind it.
Not enough data yet$0.00$0.00Not published by the provider
41.2
Poor value (41.2). Costs more per task than its benchmark results predict. Little data behind it, so treat it as a rough estimate.
Not measuredNot measured
Qwen2.5 VL 72B Instruct
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.
108.8
From 97 benchmarks. Plenty of data behind it.
Not enough data yet$0.80cache read $0.40$1.00128K
41.2
Poor value (41.2). Costs more per task than its benchmark results predict. Little data behind it, so treat it as a rough estimate.
Not measuredNot measured
Qwen2.5 VL 72B Instruct
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
108.8
From 97 benchmarks. Plenty of data behind it.
Not enough data yet$0.80cache read $0.40$1.00128K
41.2
Poor value (41.2). Costs more per task than its benchmark results predict. Little data behind it, so treat it as a rough estimate.
Not measuredNot measured

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

How this data works