Quality against price
How capable a model is, by the Epoch Capabilities Index (ECI), set against what it costs at its cheapest direct provider. Models on the frontier are the ones no other model beats on both: nothing scored higher for less.
151 of 2,612 models have an Epoch Capabilities Index; others are not evaluated. 146 of those also have a headline price and are shown below; 5 have no comparable direct price and are listed at the end. Data generated 2026-10-20 03:37 UTC.
Capabilities index against blended price
Epoch Capabilities Index, retrieved 2026-10-19. Price: each model's cheapest headline direct price, blended 3 input : 1 output.
Scatter of 146 models: Epoch Capabilities Index against blended price on a log scale. 8 models form the frontier: Mistral Nemo ($0.0178, 118.7), GPT OSS 20B ($0.0284, 137.8), Qwen3.7 Flash ($0.0518, 144.6), DeepSeek V4 Flash Latest ($0.0563, 154.5), DeepSeek V4.1 Flash ($0.143, 155.6), Claude Haiku 5.5 ($0.20, 159), GPT-6.1 Sol ($4.00, 166.3), Claude Opus 5.5 ($8.00, 166.8). All values are in the table below.
Show as table
| Model | Epoch Capabilities Index | CI ± | Blended $/M | Frontier |
|---|---|---|---|---|
| Claude Opus 5.5 | 166.8 | 3.7 | $8.00 | yes |
| GPT-6 Astra | 166.4 | 3.9 | $20 | |
| GPT-6.1 Sol | 166.3 | 3.8 | $4.00 | yes |
| Claude Fable 5.1 | 164.7 | 3.3 | $20 | |
| Gemini 4 Argon | 164.6 | 3.5 | $4.00 | |
| Claude Sonnet 5.5 | 164.5 | 3.5 | $4.00 | |
| Claude Opus 5 | 162.6 | 2.9 | $10 | |
| GPT-6 Sol | 162.6 | 2.9 | $4.00 | |
| Claude Fable 5 | 162.1 | 3.4 | $20 | |
| GPT-5.5 Pro | 162 | 3.1 | $67.50 | |
| GPT-5.6 Sol | 161.7 | 2.8 | $5.00 | |
| GPT-5.6 Terra | 159.4 | 2.7 | $4.50 | |
| GPT-5.4 Pro | 159 | 2.6 | $67.50 | |
| Claude Haiku 5.5 | 159 | 2.7 | $0.20 | yes |
| GPT-5.5 | 159 | 2.8 | $11.25 | |
| Claude Opus 4.8 | 158.3 | 2.3 | $10 | |
| Kimi K3 | 157.4 | 2.5 | $3.76 | |
| Gemini 3.7 Flash | 157.2 | 2.2 | $1.50 | |
| Gemini 3.8 Flash | 157.2 | 2.7 | $1.50 | |
| Muse Spark 1.3 | 156.8 | 2 | $2.00 | |
| GPT-5.3 Codex | 156.8 | 3.8 | $4.81 | |
| GPT-5.4 | 156.7 | 2 | $5.63 | |
| Claude Sonnet 5 | 156.4 | 2.5 | $4.00 | |
| GPT-5.6 Luna | 156.3 | 2.6 | $0.45 | |
| Claude Opus 4.7 | 156.3 | 2 | $9.00 | |
| GPT-6 Luna | 156.3 | 2.4 | $0.20 | |
| Grok 4.6 | 156.3 | 2.1 | $3.00 | |
| GPT-5.2 Pro | 156 | 2.9 | $57.75 | |
| GLM-5.3 | 155.9 | 2.2 | $1.29 | |
| DeepSeek V4.1 Flash | 155.6 | 2 | $0.143 | yes |
| DeepSeek V4 Pro 0813 | 155.3 | 2 | $1.34 | |
| Claude Opus 4.6 | 155.3 | 1.7 | $10 | |
| Muse Spark 1.2 Contributor | 154.9 | 2 | $2.00 | |
| DeepSeek V4 Flash Latest | 154.5 | 2.1 | $0.0563 | yes |
| Gemini 3.5 Flash | 154.4 | 2 | $3.38 | |
| Grok 4.7 | 154.4 | 1.9 | $2.40 | |
| Gemini 3.6 Flash | 154.2 | 1.7 | $1.50 | |
| Muse Spark 1.1 | 154 | 2.3 | $2.00 | |
| Grok 4.5 | 153.9 | 1.7 | $3.00 | |
| Qwen3.7 Max | 153.8 | 2.1 | $3.75 | |
| GPT-5.2 | 153.5 | 1.8 | $3.30 | |
| GLM-5.3-Flash | 153.2 | 2.2 | $0.162 | |
| Gemini-3-Pro | 152.9 | 2 | $4.50 | |
| Claude Sonnet 4.6 | 152.4 | 2.3 | $6.00 | |
| Grok 4.20 | 152.1 | 2.3 | $1.56 | |
| GLM-5.2 | 151.9 | 2 | $1.07 | |
| Gemini 3 Flash | 151.8 | 1.4 | $1.41 | |
| Kimi K2.6 | 150.9 | 1.8 | $0.825 | |
| GPT-5 Pro | 150.7 | 2.5 | $41.25 | |
| Inkling Small | 150.2 | 2.4 | $0.638 | |
| Claude Opus 4.5 | 150.2 | 2.4 | $10 | |
| GPT-5 | 150 | — | $3.44 | |
| GLM-5.1 | 150 | 1.9 | $1.48 | |
| Kimi K2.7 Code HighSpeed | 149.9 | 1.9 | $1.28 | |
| GPT-5.1 | 149.8 | 1.5 | $3.44 | |
| DeepSeek V4 Pro | 149.2 | 1.7 | $1.21 | |
| GPT-5.4 mini | 148.8 | 1.7 | $1.69 | |
| Inkling | 148.6 | 2.2 | $1.73 | |
| Moonshot Kimi K2.5 | 148 | 1.5 | $0.90 | |
| o3-pro | 147.4 | 2.2 | $35 | |
| Qwen3.7 Plus | 147.3 | 1.6 | $0.56 | |
| MiniMax-M3 | 147.1 | 3.6 | $0.375 | |
| Mistral Large 4 | 146.9 | 3.6 | $1.03 | |
| o3 | 146.9 | 1.9 | $3.50 | |
| Claude Sonnet 4.5 | 146.8 | 1.6 | $6.00 | |
| MiniMax-M2.5 | 146.7 | 2.7 | $0.385 | |
| Qwen3.5 397B-A17B | 146.7 | 1.7 | $0.387 | |
| Qwen3.6 27B | 146.5 | 1.8 | $0.588 | |
| Grok 4 | 146.4 | 1.8 | $6.00 | |
| GPT-5.4 nano | 146.1 | 2.4 | $0.463 | |
| deepseek-v3.2 | 146.1 | 1.4 | $0.223 | |
| DeepSeek V4 Flash | 146.1 | 2.2 | $0.0875 | |
| Moonshot Kimi K2 Thinking | 146 | 2 | $0.975 | |
| MiniMax-M2.7 | 145.9 | 4.8 | $0.518 | |
| GLM-5 | 145.8 | 1.9 | $0.738 | |
| o4-mini | 145.6 | 2 | $1.75 | |
| GPT-5 Mini | 145.4 | 1.7 | $0.525 | |
| Gemini 3.5 Flash Lite | 145.2 | 2 | $0.85 | |
| DeepSeek V3.2 Exp | 145 | 2.3 | $0.303 | |
| Qwen3.7 Flash | 144.6 | 2.3 | $0.0518 | yes |
| Gemini 3.1 Flash Lite | 144.5 | 1.9 | $0.563 | |
| Grok-4-Fast | 144.2 | 2.2 | $10 | |
| Qwen3.6 35B-A3B | 144.1 | 2.4 | $0.313 | |
| Claude Opus 4.1 | 144.1 | 2.1 | $30 | |
| GLM-4.7 | 143.8 | 2.2 | $0.738 | |
| Gemma-4-31B-IT | 142.9 | 2.4 | $0.15 | |
| Claude Opus 4 | 142.7 | 1.8 | $10 | |
| Qwen3.5 35B-A3B | 142.5 | 2.4 | $0.355 | |
| Claude Haiku 4.5 | 142.4 | 2.4 | $1.38 | |
| Qwen3 Max | 142.3 | 2.2 | $0.628 | |
| o1 | 141.8 | 1.7 | $26.25 | |
| gemma-4-26b-a4b | 141.8 | 2.4 | $0.198 | |
| Claude Sonnet 4 | 141.7 | 2.2 | $6.00 | |
| Mistral Medium 3.5 | 141.5 | 2.6 | $3.00 | |
| DeepSeek R1 0528 | 141.3 | 2.1 | $0.913 | |
| Claude Sonnet 3.7 | 141.1 | 2.2 | $6.00 | |
| Grok 3 Mini | 140.3 | 2.3 | $0.35 | |
| o3-mini | 140.3 | 2.3 | $1.93 | |
| GPT-OSS 120B | 140 | 3.2 | $0.06 | |
| DeepSeek V3.1 | 139.9 | 3.7 | $0.32 | |
| Qwen 3.5 9B (Q4_K_M) | 139.4 | 2.5 | $0.0675 | |
| Qwen3 235B-A22B | 139.4 | 2.8 | $0.27 | |
| GPT-5 Nano | 139.3 | 3.3 | $0.138 | |
| DeepSeek R1 | 139 | 2.1 | $0.92 | |
| Qwen3 32B | 138.5 | 2.8 | $0.118 | |
| Grok 3 | 138.3 | 2.1 | $6.00 | |
| Qwen3 14B | 138.2 | 3.1 | $0.12 | |
| GPT OSS 20B | 137.8 | 3.3 | $0.0284 | yes |
| QwQ 32B | 137.6 | 4.7 | $0.185 | |
| DeepSeek R1 Distill Qwen 32B | 137.4 | 4.7 | $0.15 | |
| GPT-4.1 | 136.8 | 2.4 | $3.50 | |
| Qwen3 8B | 136.2 | 3.7 | $0.06 | |
| o1 mini | 135.8 | 2.6 | $1.93 | |
| DeepSeek R1 Distill Qwen 14B | 135.4 | 5.2 | $0.07 | |
| GPT-4.1 mini | 135 | 2.7 | $0.70 | |
| Mistral Medium 3 | 134.1 | 2.8 | $0.80 | |
| DeepSeek V3 | 132.3 | 4.2 | $0.238 | |
| Llama 4 Maverick | 132.2 | 3.2 | $0.405 | |
| Claude Sonnet 3.5 | 130 | — | $6.00 | |
| GPT-4.1 nano | 129.6 | 4.3 | $0.125 | |
| qwen2-5-72b | 129 | 3.6 | $0.90 | |
| qwen2-5-32b | 128.5 | 3.8 | $0.90 | |
| Llama 3.3 70B | 127.3 | 3.9 | $0.275 | |
| Claude Haiku 3.5 | 127.2 | 4.8 | $1.60 | |
| Claude Opus 3 | 126.9 | 4 | $30 | |
| GPT-4o mini | 126.6 | 4.5 | $0.263 | |
| Nova Pro | 123.8 | 8.7 | $0.953 | |
| Gemma 3 12B | 123.5 | 6.6 | $0.238 | |
| llama-3-70b | 122.9 | 4.4 | $1.18 | |
| mixtral-8x22b | 122.1 | 4.7 | $1.20 | |
| Mistral Large | 122 | 5.3 | $0.75 | |
| Claude 3 Sonnet | 120.7 | 5.7 | $6.00 | |
| Mistral Nemo | 118.7 | 5.3 | $0.0178 | yes |
| Mixtral 8x7B | 118.5 | 5.2 | $0.50 | |
| Claude Haiku 3 | 118.4 | 5.7 | $0.50 | |
| Ministral 3B | 118.1 | 8.3 | $0.04 | |
| llama-3-8b | 116.5 | 5.8 | $0.10 | |
| llama-2-70b | 113.8 | 6.2 | $1.18 | |
| mistral-7b-v0-1 | 112.3 | 6.6 | $0.10 | |
| gemma-7b | 112.1 | 7 | $0.20 | |
| mistral-7b-v0-3 | 109.1 | 7.7 | $0.113 | |
| phi-2 | 108 | 14.5 | $0.10 | |
| llama-2-13b | 106.3 | 7.4 | $0.20 | |
| llama-2-7b | 99.2 | 8.6 | $0.10 | |
| gemma-2b | 94.3 | 9.2 | $0.10 | |
| redpajama-incite-7b-base | 90.6 | 11.3 | $0.20 |
All scored models with a price
Highest index first. Sort by any column. ± is half the width of Epoch's published confidence interval; overlapping intervals mean the order between those models is not settled.
| Model | ECI | Cheapest blended /1M | Cheapest at |
|---|---|---|---|
| Claude Opus 5.5Anthropic | 166.8 ± 3.7 | $8.00 | Anthropic |
| GPT-6 AstraOpenAI | 166.4 ± 3.9 | $20 | OpenAI |
| GPT-6.1 SolOpenAI | 166.3 ± 3.8 | $4.00 | OpenAI |
| Claude Fable 5.1Anthropic | 164.7 ± 3.3 | $20 | Anthropic |
| Gemini 4 ArgonGoogle | 164.6 ± 3.5 | $4.00 | |
| Claude Sonnet 5.5Anthropic | 164.5 ± 3.5 | $4.00 | Anthropic |
| Claude Opus 5Anthropic | 162.6 ± 2.9 | $10 | Anthropic |
| GPT-6 SolOpenAI | 162.6 ± 2.9 | $4.00 | OpenAI |
| Claude Fable 5Anthropic | 162.1 ± 3.4 | $20 | Anthropic |
| GPT-5.5 ProOpenAI | 162 ± 3.1 | $67.50 | OpenAI |
| GPT-5.6 SolOpenAI | 161.7 ± 2.8 | $5.00 | Infer by Flow7 |
| GPT-5.6 TerraOpenAI | 159.4 ± 2.7 | $4.50 | OpenAI |
| GPT-5.4 ProOpenAI | 159 ± 2.6 | $67.50 | OpenAI |
| Claude Haiku 5.5Anthropic | 159 ± 2.7 | $0.20 | Anthropic |
| GPT-5.5OpenAI | 159 ± 2.8 | $11.25 | OpenAI |
| Claude Opus 4.8Anthropic | 158.3 ± 2.3 | $10 | Anthropic |
| Kimi K3Moonshot AI | 157.4 ± 2.5 | $3.76 | Chutes |
| Gemini 3.7 FlashGoogle | 157.2 ± 2.2 | $1.50 | |
| Gemini 3.8 FlashGoogle | 157.2 ± 2.7 | $1.50 | |
| Muse Spark 1.3Meta | 156.8 ± 2 | $2.00 | Meta |
| GPT-5.3 CodexOpenAI | 156.8 ± 3.8 | $4.81 | OpenAI |
| GPT-5.4OpenAI | 156.7 ± 2 | $5.63 | OpenAI |
| Claude Sonnet 5Anthropic | 156.4 ± 2.5 | $4.00 | Anthropic |
| GPT-5.6 LunaOpenAI | 156.3 ± 2.6 | $0.45 | OpenAI |
| Claude Opus 4.7Anthropic | 156.3 ± 2 | $9.00 | GMI Cloud |
| GPT-6 LunaOpenAI | 156.3 ± 2.4 | $0.20 | OpenAI |
| Grok 4.6xAI | 156.3 ± 2.1 | $3.00 | xAI |
| GPT-5.2 ProOpenAI | 156 ± 2.9 | $57.75 | OpenAI |
| GLM-5.3Z.ai (Zhipu AI) | 155.9 ± 2.2 | $1.29 | Phala |
| DeepSeek V4.1 FlashDeepSeek | 155.6 ± 2 | $0.143 | sail |
| DeepSeek V4 Pro 0813DeepSeek | 155.3 ± 2 | $1.34 | IteraCompute |
| Claude Opus 4.6Anthropic | 155.3 ± 1.7 | $10 | Anthropic |
| Muse Spark 1.2 ContributorMeta | 154.9 ± 2 | $2.00 | Meta |
| DeepSeek V4 Flash LatestDeepSeek | 154.5 ± 2.1 | $0.0563 | engy |
| Gemini 3.5 FlashGoogle | 154.4 ± 2 | $3.38 | |
| Grok 4.7xAI | 154.4 ± 1.9 | $2.40 | Prime Inference |
| Gemini 3.6 FlashGoogle | 154.2 ± 1.7 | $1.50 | |
| Muse Spark 1.1Meta | 154 ± 2.3 | $2.00 | Meta |
| Grok 4.5xAI | 153.9 ± 1.7 | $3.00 | xAI |
| Qwen3.7 MaxAlibaba (Qwen) | 153.8 ± 2.1 | $3.75 | Alibaba |
| GPT-5.2OpenAI | 153.5 ± 1.8 | $3.30 | SAP AI Core |
| GLM-5.3-FlashZ.ai (Zhipu AI) | 153.2 ± 2.2 | $0.162 | IteraCompute |
| Gemini-3-ProGoogle | 152.9 ± 2 | $4.50 | replicate |
| Claude Sonnet 4.6Anthropic | 152.4 ± 2.3 | $6.00 | Anthropic |
| Grok 4.20xAI | 152.1 ± 2.3 | $1.56 | xAI |
| GLM-5.2Z.ai (Zhipu AI) | 151.9 ± 2 | $1.07 | CoreWeave |
| Gemini 3 FlashGoogle | 151.8 ± 1.4 | $1.41 | Databricks |
| Kimi K2.6Moonshot AI | 150.9 ± 1.8 | $0.825 | Chutes |
| GPT-5 ProOpenAI | 150.7 ± 2.5 | $41.25 | OpenAI |
| Inkling SmallThinking Machines | 150.2 ± 2.4 | $0.638 | Deep Infra |
| Claude Opus 4.5Anthropic | 150.2 ± 2.4 | $10 | Anthropic |
| GPT-5OpenAI | 150 | $3.44 | OpenAI |
| GLM-5.1Z.ai (Zhipu AI) | 150 ± 1.9 | $1.48 | Baidu |
| Kimi K2.7 Code HighSpeedMoonshot AI | 149.9 ± 1.9 | $1.28 | StreamLake |
| GPT-5.1OpenAI | 149.8 ± 1.5 | $3.44 | OpenAI |
| DeepSeek V4 ProDeepSeek | 149.2 ± 1.7 | $1.21 | Parasail |
| GPT-5.4 miniOpenAI | 148.8 ± 1.7 | $1.69 | OpenAI |
| InklingThinking Machines | 148.6 ± 2.2 | $1.73 | Deep Infra |
| Moonshot Kimi K2.5Moonshot AI | 148 ± 1.5 | $0.90 | SiliconFlow |
| o3-proOpenAI | 147.4 ± 2.2 | $35 | OpenAI |
| Qwen3.7 PlusAlibaba (Qwen) | 147.3 ± 1.6 | $0.56 | Together AI |
| MiniMax-M3MiniMax | 147.1 ± 3.6 | $0.375 | Vultr |
| Mistral Large 4Mistral AI | 146.9 ± 3.6 | $1.03 | Mistral |
| o3OpenAI | 146.9 ± 1.9 | $3.50 | OpenAI |
| Claude Sonnet 4.5Anthropic | 146.8 ± 1.6 | $6.00 | Amazon Bedrock |
| MiniMax-M2.5MiniMax | 146.7 ± 2.7 | $0.385 | DInference |
| Qwen3.5 397B-A17BAlibaba (Qwen) | 146.7 ± 1.7 | $0.387 | Alibaba (China) |
| Qwen3.6 27BAlibaba (Qwen) | 146.5 ± 1.8 | $0.588 | STACKIT |
| Grok 4xAI | 146.4 ± 1.8 | $6.00 | Azure |
| GPT-5.4 nanoOpenAI | 146.1 ± 2.4 | $0.463 | OpenAI |
| deepseek-v3.2DeepSeek | 146.1 ± 1.4 | $0.223 | CrofAI |
| DeepSeek V4 FlashDeepSeek | 146.1 ± 2.2 | $0.0875 | StreamLake |
| Moonshot Kimi K2 ThinkingMoonshot AI | 146 ± 2 | $0.975 | IO.NET |
| MiniMax-M2.7MiniMax | 145.9 ± 4.8 | $0.518 | Moark |
| GLM-5Z.ai (Zhipu AI) | 145.8 ± 1.9 | $0.738 | Vultr |
| o4-miniOpenAI | 145.6 ± 2 | $1.75 | replicate |
| GPT-5 MiniOpenAI | 145.4 ± 1.7 | $0.525 | Snowflake Cortex |
| Gemini 3.5 Flash LiteGoogle | 145.2 ± 2 | $0.85 | |
| DeepSeek V3.2 ExpDeepSeek | 145 ± 2.3 | $0.303 | Meganova |
| Qwen3.7 FlashAlibaba (Qwen) | 144.6 ± 2.3 | $0.0518 | Alibaba (China) |
| Gemini 3.1 Flash LiteGoogle | 144.5 ± 1.9 | $0.563 | |
| Grok-4-FastxAI | 144.2 ± 2.2 | $10 | OCI Generative AI |
| Qwen3.6 35B-A3BAlibaba (Qwen) | 144.1 ± 2.4 | $0.313 | Deep Infra |
| Claude Opus 4.1Anthropic | 144.1 ± 2.1 | $30 | Databricks |
| GLM-4.7Z.ai (Zhipu AI) | 143.8 ± 2.2 | $0.738 | Deep Infra |
| Gemma-4-31B-ITGoogle | 142.9 ± 2.4 | $0.15 | CrofAI |
| Claude Opus 4Anthropic | 142.7 ± 1.8 | $10 | Snowflake Cortex |
| Qwen3.5 35B-A3BAlibaba (Qwen) | 142.5 ± 2.4 | $0.355 | Deep Infra |
| Claude Haiku 4.5Anthropic | 142.4 ± 2.4 | $1.38 | Xpersona |
| Qwen3 MaxAlibaba (Qwen) | 142.3 ± 2.2 | $0.628 | Alibaba (China) |
| o1OpenAI | 141.8 ± 1.7 | $26.25 | DigitalOcean |
| gemma-4-26b-a4bGoogle | 141.8 ± 2.4 | $0.198 | Amazon Bedrock (mantle) |
| Claude Sonnet 4Anthropic | 141.7 ± 2.2 | $6.00 | DigitalOcean |
| Mistral Medium 3.5Mistral AI | 141.5 ± 2.6 | $3.00 | Mistral |
| DeepSeek R1 0528DeepSeek | 141.3 ± 2.1 | $0.913 | Deep Infra |
| Claude Sonnet 3.7Anthropic | 141.1 ± 2.2 | $6.00 | Amazon Bedrock |
| Grok 3 MinixAI | 140.3 ± 2.3 | $0.35 | xAI |
| o3-miniOpenAI | 140.3 ± 2.3 | $1.93 | DigitalOcean |
| GPT-OSS 120BOpenAI | 140 ± 3.2 | $0.06 | NEAR AI Cloud |
| DeepSeek V3.1DeepSeek | 139.9 ± 3.7 | $0.32 | TokenGo |
| Qwen 3.5 9B (Q4_K_M)Alibaba (Qwen) | 139.4 ± 2.5 | $0.0675 | CrofAI |
| Qwen3 235B-A22BAlibaba (Qwen) | 139.4 ± 2.8 | $0.27 | Deep Infra |
| GPT-5 NanoOpenAI | 139.3 ± 3.3 | $0.138 | OpenAI |
| DeepSeek R1DeepSeek | 139 ± 2.1 | $0.92 | SiliconFlow |
| Qwen3 32BAlibaba (Qwen) | 138.5 ± 2.8 | $0.118 | OVHcloud AI Endpoints |
| Grok 3xAI | 138.3 ± 2.1 | $6.00 | xAI |
| Qwen3 14BAlibaba (Qwen) | 138.2 ± 3.1 | $0.12 | Nebius Token Factory |
| GPT OSS 20BOpenAI | 137.8 ± 3.3 | $0.0284 | Darkbloom |
| QwQ 32BAlibaba (Qwen) | 137.6 ± 4.7 | $0.185 | nscale |
| DeepSeek R1 Distill Qwen 32BDeepSeek | 137.4 ± 4.7 | $0.15 | nscale |
| GPT-4.1OpenAI | 136.8 ± 2.4 | $3.50 | OpenAI |
| Qwen3 8BAlibaba (Qwen) | 136.2 ± 3.7 | $0.06 | SiliconFlow |
| o1 miniOpenAI | 135.8 ± 2.6 | $1.93 | OpenAI |
| DeepSeek R1 Distill Qwen 14BDeepSeek | 135.4 ± 5.2 | $0.07 | nscale |
| GPT-4.1 miniOpenAI | 135 ± 2.7 | $0.70 | OpenAI |
| Mistral Medium 3Mistral AI | 134.1 ± 2.8 | $0.80 | Vertex |
| DeepSeek V3DeepSeek | 132.3 ± 4.2 | $0.238 | inferenceapis |
| Llama 4 MaverickMeta | 132.2 ± 3.2 | $0.405 | DigitalOcean |
| Claude Sonnet 3.5Anthropic | 130 | $6.00 | Amazon Bedrock |
| GPT-4.1 nanoOpenAI | 129.6 ± 4.3 | $0.125 | SAP AI Core |
| qwen2-5-72bAlibaba (Qwen) | 129 ± 3.6 | $0.90 | Fireworks AI |
| qwen2-5-32bAlibaba (Qwen) | 128.5 ± 3.8 | $0.90 | Fireworks AI |
| Llama 3.3 70BMeta | 127.3 ± 3.9 | $0.275 | inferenceapis |
| Claude Haiku 3.5Anthropic | 127.2 ± 4.8 | $1.60 | Amazon Bedrock |
| Claude Opus 3Anthropic | 126.9 ± 4 | $30 | Amazon Bedrock |
| GPT-4o miniOpenAI | 126.6 ± 4.5 | $0.263 | OpenAI |
| Nova ProAmazon | 123.8 ± 8.7 | $0.953 | SAP AI Core |
| Gemma 3 12BGoogle | 123.5 ± 6.6 | $0.238 | Neon |
| llama-3-70bMeta | 122.9 ± 4.4 | $1.18 | replicate |
| mixtral-8x22bMistral AI | 122.1 ± 4.7 | $1.20 | Fireworks AI |
| Mistral LargeMistral AI | 122 ± 5.3 | $0.75 | Mistral |
| Claude 3 SonnetAnthropic | 120.7 ± 5.7 | $6.00 | Vertex (Anthropic) |
| Mistral NemoMistral AI | 118.7 ± 5.3 | $0.0178 | IO.NET |
| Mixtral 8x7BMistral AI | 118.5 ± 5.2 | $0.50 | Fireworks AI |
| Claude Haiku 3Anthropic | 118.4 ± 5.7 | $0.50 | SAP AI Core |
| Ministral 3BMistral AI | 118.1 ± 8.3 | $0.04 | Mistral |
| llama-3-8bMeta | 116.5 ± 5.8 | $0.10 | replicate |
| llama-2-70bMeta | 113.8 ± 6.2 | $1.18 | replicate |
| mistral-7b-v0-1Mistral AI | 112.3 ± 6.6 | $0.10 | replicate |
| gemma-7bGoogle | 112.1 ± 7 | $0.20 | Fireworks AI |
| mistral-7b-v0-3Mistral AI | 109.1 ± 7.7 | $0.113 | llamagate |
| phi-2Microsoft | 108 ± 14.5 | $0.10 | Together AI |
| llama-2-13bMeta | 106.3 ± 7.4 | $0.20 | replicate |
| llama-2-7bMeta | 99.2 ± 8.6 | $0.10 | replicate |
| gemma-2bGoogle | 94.3 ± 9.2 | $0.10 | Together AI |
| redpajama-incite-7b-baseTogether AI | 90.6 ± 11.3 | $0.20 | Together AI |
Evaluated, but no headline-eligible direct price, so not on the chart: o1-preview (134.8 ± 3.7), llama-3-2-90b (125.5 ± 3.9), Claude 2.0 / 2.1 (120.2 ± 5.1), gemma-2-9b (119.9 ± 5.5), Llama-3.2-1B (102 ± 12.3).
Method and licence
Index. The Epoch Capabilities Index is Epoch AI's combined score across its own benchmark runs, published with a confidence interval. It is an index, not a percentage, so dividing it by price would produce a number with no meaning. This page never does that; it shows both and marks the frontier.
Price. The same headline price as the rest of the site: the cheapest headline-eligible direct endpoint (no gateways, no free tiers, no prices the sanity check flags), blended 3 input : 1 output per 1M tokens. The x-axis is logarithmic because prices span three orders of magnitude.
Matching. Epoch's model names are matched to catalogue models with the same matcher and hand-checked pins as the Epoch model facts (release dates, parameters) elsewhere on the site.
Not evaluated is not zero. A model without an index is not ranked and not plotted. Speed and latency are not covered: Epoch does not publish them.
Licence. Only Epoch's own runs and the ECI are used, under CC BY 4.0. Benchmark files Epoch hosts for other publishers keep their own licences and are not used.
Epoch AI, 'Capabilities & benchmarking'. Published online at epoch.ai. Retrieved from 'https://epoch.ai/benchmarks/use-this-data' [online resource]. Accessed 19 Oct 2026.