Skip to content

Listings

All listings: Parasail

Every listing for the models on page 2 of 3 of Parasail, including each model's other versions and routes. Back to Parasail

Sold by Parasail

22 listings on this page sold by Parasail.

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
GPT OSS 120B
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.
parasail-gpt-oss-120b-fast
118.2
From 424 benchmarks. Plenty of data behind it.
86Β Β±5
Coding IQ 85.5, 95% interval 80–91. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.15$0.75131K
58.4
Good value (58.4). Cheaper per task than its benchmark results predict. Plenty of data behind it.
37 Β±11
Terse (37 Β±11). Generates fewer tokens than models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
140.2
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GPT OSS 120B
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.
parasail-gpt-oss-120b
118.2
From 424 benchmarks. Plenty of data behind it.
86Β Β±5
Coding IQ 85.5, 95% interval 80–91. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.10cache read $0.055$0.75131K
58.4
Good value (58.4). Cheaper per task than its benchmark results predict. Plenty of data behind it.
37 Β±11
Terse (37 Β±11). Generates fewer tokens than models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
140.2
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Qwen3 Coder Next
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.
parasail-qwen3-coder-next Β· bf16
118.2
From 27 benchmarks. Plenty of data behind it.
98Β Β±7
Coding IQ 97.7, 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.
$0.12cache read $0.07$0.80262K
36.8
Poor value (36.8). Costs more per task than its benchmark results predict. Plenty of data behind it.
54 Β±9
Typical (54 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
Qwen3 Next 80B A3B 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.
parasail-qwen-3-next-80b-instruct Β· fp8
118.1
From 154 benchmarks. Plenty of data behind it.
92Β Β±11
Coding IQ 91.7, 95% interval 81–102. A fair amount of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.10cache read $0.07$1.10262K
47.6
Fair value (47.6). Costs per task about what its benchmark results predict. A fair amount of data behind it.
Not measuredNot measured
Mistral Small 3.2 24B Instruct
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.
parasail-mistral-small-32-24b Β· 2506
113.3
From 50 benchmarks. Plenty of data behind it.
Not enough data yet$0.09cache read $0.05$0.30131K
58.5
Good value (58.5). Cheaper per task than its benchmark results predict. Little data behind it, so treat it as a rough estimate.
Not measuredNot measured
Qwen3.5 9B
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.
parasail-qwen35-9b Β· bf16
113.2
From 223 benchmarks. Plenty of data behind it.
86Β Β±8
Coding IQ 85.6, 95% interval 77–94. A fair amount of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.10$0.25262K
55.4
Good value (55.4). Cheaper per task than its benchmark results predict. Plenty of data behind it.
Not measured
139.5
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Llama 4 Maverick
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.
parasail-llama-4-maverick-instruct-fp8 Β· fp8
111.8
From 350 benchmarks. Plenty of data behind it.
78Β Β±9
Coding IQ 78.1, 95% interval 69–87. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.35cache read $0.17$1.00524K
42.1
Poor value (42.1). Costs more per task than its benchmark results predict. Plenty of data behind it.
38 Β±9
Terse (38 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.
132.2
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.
parasail-gpt-oss-20b Β· fp4
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
Qwen3 VL 8B 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.
parasail-qwen3vl-8b-instruct
104.9
From 81 benchmarks. Plenty of data behind it.
Not enough data yet$0.25cache read $0.12$0.75262K
27.3
Poor value (27.3). Costs more per task than its benchmark results predict. A fair amount of data behind it.
Not measuredNot measured
BGE M3
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.
parasail-bge-m3
100.9
From 136 benchmarks. Plenty of data behind it.
Not enough data yet$0.01β€”8KNot enough data yetNot measuredNot measured
Gemma 3 4B IT
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.
parasail-googlegemma-3-4b-it
95.2
From 249 benchmarks. Plenty of data behind it.
67Β Β±19
Coding IQ 66.8, 95% interval 48–86. A fair amount of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.04cache read $0.02$0.05131K
38.3
Poor value (38.3). Costs more per task than its benchmark results predict. Little data behind it, so treat it as a rough estimate.
Not measuredNot measured
L3 8B Lunaris V1parasail-sao10kl3-lunaris-8b
93.9
From 6 benchmarks. A fair amount of data behind it.
Not enough data yet$0.04$0.058KNot enough data yetNot measuredNot measured
Llama 3.2 3B
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.
parasail-llama-32-3b-instruct Β· bf16
86.5
From 255 benchmarks. Plenty of data behind it.
β‰ˆ 76Β Β±14
Coding IQ 75.7, 95% interval 62–89. 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.
$0.05$0.33131K
44.3
Poor value (44.3). Costs more per task than its benchmark results predict. Little data behind it, so treat it as a rough estimate.
Not measuredNot measured
Mistral Nemo Instruct 2407 FP8parasail-mistralaimistral-nemo Β· fp8
Not enough data yetNot enough data yet$0.03$0.03131KNot enough data yetNot measuredNot measured
Gemma 3 27b It FP8 Dynamicparasail-gemma3-27b-it
Not enough data yetNot enough data yet$0.08cache read $0.04$0.45131KNot enough data yetNot measuredNot measured
MythoMax L2 13b
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
parasail-mythomax-13b
Not enough data yetNot enough data yet$0.08$0.114KNot enough data yetNot measuredNot measured
UI TARS 1.5 7Bparasail-ui-tars-1p5-7b
Not enough data yetNot enough data yet$0.10$0.20128KNot enough data yetNot measuredNot measured
DeepSeek V4 Flash DSpark
πŸ‡¨πŸ‡³
Maker headquartered in CN, according to the company's own website. The citation is on the model page.
parasail-deepseek-v4-flash
Not enough data yetNot enough data yet$0.14cache read $0.07$0.281MNot enough data yetNot measuredNot measured
Gemma 4 31B It FP8 Dynamicparasail-gemma-4-31b-it
Not enough data yetNot enough data yet$0.15cache read $0.06$0.40262KNot enough data yetNot measuredNot measured
GLM 5.3 Flash NVFP4parasail-glm-53-flash-fast Β· nvfp4
Not enough data yetNot enough data yet$0.1875cache read $0.0375$0.6251MNot enough data yetNot measuredNot measured
Llama 3.3 70B Instruct FP8 Dynamicparasail-llama-33-70b-fp8 Β· fp8
Not enough data yetNot enough data yet$0.22cache read $0.11$0.50131KNot enough data yetNot measuredNot measured
Cydonia 24B V4.1
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
Newer version
Newer versionCydonia-24B-v4.3We compared release dates and version names in our catalogue.
parasail-cydonia-24-v41
Not enough data yetNot enough data yet$0.30cache read $0.15$0.50131K
55.4
Good value (55.4). Cheaper per task than its benchmark results predict. Little data behind it, so treat it as a rough estimate.
Not measuredNot measured

Run by Parasail

43 listings on this page where Parasail runs the model. Another provider sells these listings.

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
ModelProviderllminfo IQCoding IQInput $/1MOutput $/1MMax Input TokensValueYapEpoch Capabilities Index
GPT OSS 120B
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.
Β Hugging Face
118.2
From 424 benchmarks. Plenty of data behind it.
86Β Β±5
Coding IQ 85.5, 95% interval 80–91. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.10$0.75131K
58.4
Good value (58.4). Cheaper per task than its benchmark results predict. Plenty of data behind it.
37 Β±11
Terse (37 Β±11). Generates fewer tokens than models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
140.2
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GPT OSS 120B
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 Β· πŸ‡ΊπŸ‡Έ US
Β NanoGPT
118.2
From 424 benchmarks. Plenty of data behind it.
86Β Β±5
Coding IQ 85.5, 95% interval 80–91. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.105cache read $0.0578$0.7875Not published by the provider
58.4
Good value (58.4). Cheaper per task than its benchmark results predict. Plenty of data behind it.
37 Β±11
Terse (37 Β±11). Generates fewer tokens than models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
140.2
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GPT OSS 120B
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.
fast Β· fp4
Β OpenRouter
118.2
From 424 benchmarks. Plenty of data behind it.
86Β Β±5
Coding IQ 85.5, 95% interval 80–91. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.15$0.75131K
58.4
Good value (58.4). Cheaper per task than its benchmark results predict. Plenty of data behind it.
37 Β±11
Terse (37 Β±11). Generates fewer tokens than models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
140.2
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
GPT OSS 120B
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
Β OpenRouter
118.2
From 424 benchmarks. Plenty of data behind it.
86Β Β±5
Coding IQ 85.5, 95% interval 80–91. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.10cache read $0.055$0.75131K
58.4
Good value (58.4). Cheaper per task than its benchmark results predict. Plenty of data behind it.
37 Β±11
Terse (37 Β±11). Generates fewer tokens than models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.
140.2
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Qwen3 Coder Next
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 Hugging Face
118.2
From 27 benchmarks. Plenty of data behind it.
98Β Β±7
Coding IQ 97.7, 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.
$0.12$0.80262K
36.8
Poor value (36.8). Costs more per task than its benchmark results predict. Plenty of data behind it.
54 Β±9
Typical (54 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
Qwen3 Coder Next
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.
bf16 Β· πŸ‡ΊπŸ‡Έ US
Β NanoGPT
118.2
From 27 benchmarks. Plenty of data behind it.
98Β Β±7
Coding IQ 97.7, 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.
$0.126cache read $0.0735$0.84Not published by the provider
36.8
Poor value (36.8). Costs more per task than its benchmark results predict. Plenty of data behind it.
54 Β±9
Typical (54 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
Qwen3 Coder Next
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.
bf16
Β OpenRouter
118.2
From 27 benchmarks. Plenty of data behind it.
98Β Β±7
Coding IQ 97.7, 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.
$0.12cache read $0.07$0.80262K
36.8
Poor value (36.8). Costs more per task than its benchmark results predict. Plenty of data behind it.
54 Β±9
Typical (54 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
Qwen3 Next 80B A3B 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 Hugging Face
118.1
From 154 benchmarks. Plenty of data behind it.
92Β Β±11
Coding IQ 91.7, 95% interval 81–102. A fair amount of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.10$1.10262K
47.6
Fair value (47.6). Costs per task about what its benchmark results predict. A fair amount of data behind it.
Not measuredNot measured
Qwen3 Next 80B A3B 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 Β· πŸ‡ΊπŸ‡Έ US
Β NanoGPT
118.1
From 154 benchmarks. Plenty of data behind it.
92Β Β±11
Coding IQ 91.7, 95% interval 81–102. A fair amount of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.105cache read $0.0735$1.155Not published by the provider
47.6
Fair value (47.6). Costs per task about what its benchmark results predict. A fair amount of data behind it.
Not measuredNot measured
Qwen3 Next 80B A3B 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
Β OpenRouter
118.1
From 154 benchmarks. Plenty of data behind it.
92Β Β±11
Coding IQ 91.7, 95% interval 81–102. A fair amount of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.10cache read $0.07$1.10262K
47.6
Fair value (47.6). Costs per task about what its benchmark results predict. A fair amount of data behind it.
Not measuredNot measured
Mistral Small 3.2 24B Instruct
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.
bf16
Β OpenRouter
113.3
From 50 benchmarks. Plenty of data behind it.
Not enough data yet$0.09cache read $0.05$0.30131K
58.5
Good value (58.5). Cheaper per task than its benchmark results predict. Little data behind it, so treat it as a rough estimate.
Not measuredNot measured
Qwen3.5 9B
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 Hugging Face
113.2
From 223 benchmarks. Plenty of data behind it.
86Β Β±8
Coding IQ 85.6, 95% interval 77–94. A fair amount of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.10$0.25262K
55.4
Good value (55.4). Cheaper per task than its benchmark results predict. Plenty of data behind it.
Not measured
139.5
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Qwen3.5 9B
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.
bf16 Β· πŸ‡ΊπŸ‡Έ US
Β NanoGPT
113.2
From 223 benchmarks. Plenty of data behind it.
86Β Β±8
Coding IQ 85.6, 95% interval 77–94. A fair amount of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.105$0.2625Not published by the provider
55.4
Good value (55.4). Cheaper per task than its benchmark results predict. Plenty of data behind it.
Not measured
139.5
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Qwen3.5 9B
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.
bf16
Β OpenRouter
113.2
From 223 benchmarks. Plenty of data behind it.
86Β Β±8
Coding IQ 85.6, 95% interval 77–94. A fair amount of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.10$0.25262K
55.4
Good value (55.4). Cheaper per task than its benchmark results predict. Plenty of data behind it.
Not measured
139.5
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Llama 4 Maverick
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.
meta-llama/Llama-4-Maverick-17B-128E-Instruct
Β Hugging Face
111.8
From 350 benchmarks. Plenty of data behind it.
78Β Β±9
Coding IQ 78.1, 95% interval 69–87. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.35$1.00524K
42.1
Poor value (42.1). Costs more per task than its benchmark results predict. Plenty of data behind it.
38 Β±9
Terse (38 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.
132.2
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Llama 4 Maverick
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 Β· πŸ‡ΊπŸ‡Έ US
Β NanoGPT
111.8
From 350 benchmarks. Plenty of data behind it.
78Β Β±9
Coding IQ 78.1, 95% interval 69–87. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.3675cache read $0.1785$1.05Not published by the provider
42.1
Poor value (42.1). Costs more per task than its benchmark results predict. Plenty of data behind it.
38 Β±9
Terse (38 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.
132.2
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Llama 4 Maverick
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
Β OpenRouter
111.8
From 350 benchmarks. Plenty of data behind it.
78Β Β±9
Coding IQ 78.1, 95% interval 69–87. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.35cache read $0.17$1.00524K
42.1
Poor value (42.1). Costs more per task than its benchmark results predict. Plenty of data behind it.
38 Β±9
Terse (38 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.
132.2
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.
Β Hugging Face
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$0.15131K
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 Β· πŸ‡ΊπŸ‡Έ US
Β NanoGPT
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.
$0.0315cache read $0.021$0.1575Not 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.
fp4
Β OpenRouter
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
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.
Β Hugging Face
108.8
From 97 benchmarks. Plenty of data behind it.
Not enough data yet$0.80$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
Β OpenRouter
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.
parasail/parasail-qwen25-vl-72b-instruct
Β Requesty
108.8
From 97 benchmarks. Plenty of data behind it.
Not enough data yet$0.70$0.7033K
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
Gemma 3 27B IT
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.
Β Hugging Face
108.5
From 365 benchmarks. Plenty of data behind it.
73Β Β±8
Coding IQ 72.7, 95% interval 65–81. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.08$0.45131K
46.2
Poor value (46.2). Costs more per task than its benchmark results predict. A fair amount of data behind it.
Not measured
127.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Gemma 3 27B IT
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
Β OpenRouter
108.5
From 365 benchmarks. Plenty of data behind it.
73Β Β±8
Coding IQ 72.7, 95% interval 65–81. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.08cache read $0.04$0.45131K
46.2
Poor value (46.2). Costs more per task than its benchmark results predict. A fair amount of data behind it.
Not measured
127.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Gemma 3 27B IT
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.
Β Requesty
108.5
From 365 benchmarks. Plenty of data behind it.
73Β Β±8
Coding IQ 72.7, 95% interval 65–81. Plenty of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.30$0.50128K
46.2
Poor value (46.2). Costs more per task than its benchmark results predict. A fair amount of data behind it.
Not measured
127.6
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Llama 3.3 70B Instruct
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.
Β Hugging Face
108.1
From 278 benchmarks. Plenty of data behind it.
77Β Β±9
Coding IQ 77.1, 95% interval 68–86. A fair amount of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.22$0.50131K
31.6
Poor value (31.6). Costs more per task than its benchmark results predict. A fair amount of data behind it.
Not measured
126.2
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Llama 3.3 70B Instruct
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 Β· πŸ‡ΊπŸ‡Έ US
Β NanoGPT
108.1
From 278 benchmarks. Plenty of data behind it.
77Β Β±9
Coding IQ 77.1, 95% interval 68–86. A fair amount of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.231cache read $0.1155$0.525Not published by the provider
31.6
Poor value (31.6). Costs more per task than its benchmark results predict. A fair amount of data behind it.
Not measured
126.2
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Llama 3.3 70B Instruct
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
Β OpenRouter
108.1
From 278 benchmarks. Plenty of data behind it.
77Β Β±9
Coding IQ 77.1, 95% interval 68–86. A fair amount of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.22cache read $0.11$0.50131K
31.6
Poor value (31.6). Costs more per task than its benchmark results predict. A fair amount of data behind it.
Not measured
126.2
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Qwen3 VL 8B 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 Hugging Face
104.9
From 81 benchmarks. Plenty of data behind it.
Not enough data yet$0.25$0.75262K
27.3
Poor value (27.3). Costs more per task than its benchmark results predict. A fair amount of data behind it.
Not measuredNot measured
Qwen3 VL 8B 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.
bf16
Β OpenRouter
104.9
From 81 benchmarks. Plenty of data behind it.
Not enough data yet$0.25cache read $0.12$0.75262K
27.3
Poor value (27.3). Costs more per task than its benchmark results predict. A fair amount of data behind it.
Not measuredNot measured
Mistral Nemo
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.
fp8
Β OpenRouter
98.1
From 85 benchmarks. Plenty of data behind it.
Not enough data yet$0.03$0.03131K
52.2
Fair value (52.2). Costs per task about what its benchmark results predict. Little data behind it, so treat it as a rough estimate.
Not measured
113.9
Epoch Capabilities IndexSource: Epoch AI Benchmarking Hub API
Gemma 3 4B IT
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.
Β Hugging Face
95.2
From 249 benchmarks. Plenty of data behind it.
67Β Β±19
Coding IQ 66.8, 95% interval 48–86. A fair amount of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.04$0.05131K
38.3
Poor value (38.3). Costs more per task than its benchmark results predict. Little data behind it, so treat it as a rough estimate.
Not measuredNot measured
Gemma 3 4B IT
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
Β OpenRouter
95.2
From 249 benchmarks. Plenty of data behind it.
67Β Β±19
Coding IQ 66.8, 95% interval 48–86. A fair amount of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.
$0.04cache read $0.02$0.05131K
38.3
Poor value (38.3). Costs more per task than its benchmark results predict. Little data behind it, so treat it as a rough estimate.
Not measuredNot measured
Β Hugging Face
93.9
From 6 benchmarks. A fair amount of data behind it.
Not enough data yet$0.04$0.058KNot enough data yetNot measuredNot measured
Llama 3.2 3B
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.
meta-llama/Llama-3.2-3B-Instruct
Β Hugging Face
86.5
From 255 benchmarks. Plenty of data behind it.
β‰ˆ 76Β Β±14
Coding IQ 75.7, 95% interval 62–89. 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.
$0.05$0.33131K
44.3
Poor value (44.3). Costs more per task than its benchmark results predict. Little data behind it, so treat it as a rough estimate.
Not measuredNot measured
Llama 3.2 3B
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.
meta-llama/llama-3.2-3b-instruct Β· bf16 Β· πŸ‡ΊπŸ‡Έ US
Β NanoGPT
86.5
From 255 benchmarks. Plenty of data behind it.
β‰ˆ 76Β Β±14
Coding IQ 75.7, 95% interval 62–89. 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.
$0.0525$0.3465Not published by the provider
44.3
Poor value (44.3). Costs more per task than its benchmark results predict. Little data behind it, so treat it as a rough estimate.
Not measuredNot measured
Llama 3.2 3B
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.
meta-llama/llama-3.2-3b-instruct Β· bf16
Β OpenRouter
86.5
From 255 benchmarks. Plenty of data behind it.
β‰ˆ 76Β Β±14
Coding IQ 75.7, 95% interval 62–89. 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.
$0.05$0.33131K
44.3
Poor value (44.3). Costs more per task than its benchmark results predict. Little data behind it, so treat it as a rough estimate.
Not measuredNot measured
MythoMax L2 13b
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
Β Hugging FaceNot enough data yetNot enough data yet$0.08$0.114KNot enough data yetNot measuredNot measured
MythoMax L2 13b
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
fp16
Β OpenRouterNot enough data yetNot enough data yet$0.08$0.114KNot enough data yetNot measuredNot measured
Β Hugging FaceNot enough data yetNot enough data yet$0.10$0.20128KNot enough data yetNot measuredNot measured
Β OpenRouterNot enough data yetNot enough data yet$0.10$0.20128KNot enough data yetNot measuredNot measured
Cydonia 24B V4.1
Open weights. The weights are publicly available, according to the provider's model list or a public Hugging Face repo.
Newer version
Newer versionCydonia-24B-v4.3We compared release dates and version names in our catalogue.
bf16
Β OpenRouterNot enough data yetNot enough data yet$0.30cache read $0.15$0.50131K
55.4
Good value (55.4). Cheaper 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