Benchmarks
AI benchmarks
Which AI models are most capable right now? This table shows independent test results from Epoch AI, a research group that runs the same tests on every model. We don't run our own tests or change their numbers.
Up to date. Data updated 1 of 1 source current Data status for all sources
Showing all 270 models, sorted by ECI.
| # | |||||||||
|---|---|---|---|---|---|---|---|---|---|
| 1 | Claude Opus 5.5 | Anthropic | 168.8[165–174] | 90.6 | 91.2 | 100.0 | – | 72.2 | |
| 2 | GPT-6 Astra | OpenAI | 168.8[165–175] | 95.8 | 93.7 | 100.0 | – | 75.6 | |
| 3 | GPT-6.1 Sol | OpenAI | 168.4[165–174] | 95.4 | 93.7 | 100.0 | – | 73.9 | |
| 4 | Claude Fable 5.1 | Anthropic | 166.5[163–171] | – | 90.2 | 100.0 | – | 70.8 | |
| 5 | Gemini 4 Argon | Google DeepMind | 166.4[163–172] | 96.4 | 86.0 | 100.0 | – | 76.0 | |
| 6 | Claude Sonnet 5.5 | Anthropic | 166.2[163–171] | 95.6 | 88.8 | 100.0 | – | 46.5 | |
| 7 | Claude Opus 5 | Anthropic | 163.9[161–168] | 93.9 | 85.6 | 98.9 | – | 59.9 | |
| 8 | GPT-6 Sol | OpenAI | 163.7[161–167] | 94.3 | 89.8 | 100.0 | – | 60.7 | |
| 9 | Claude Fable 5 | Anthropic | 163.4[161–168] | 85.9 | 87.0 | 100.0 | – | 70.7 | |
| 10 | GPT-5.5 Pro | OpenAI | 162.9[160–167] | 93.9 | 87.7 | 100.0 | – | – | |
| 11 | GPT-5.6 Sol | OpenAI | 162.8[160–167] | 93.5 | 89.1 | 100.0 | – | 69.7 | |
| 12 | Muse Wonder | Meta AI | 161.8[159–165] | – | 82.1 | 99.4 | – | – | |
| 13 | GPT-5.6 Terra | OpenAI | 160.1[157–163] | 93.3 | 86.0 | 99.7 | – | 43.2 | |
| 14 | GPT-5.5 | OpenAI | 159.7[157–164] | 94.0 | 85.3 | 100.0 | 80.6 | 63.0 | |
| 15 | GPT-5.4 Pro | OpenAI | 159.6[157–163] | 94.6 | 82.5 | – | – | 46.3 | |
| 16 | Claude Haiku 5.5 | Anthropic | 159.3[156–162] | 89.6 | 75.1 | 98.9 | – | 23.8 | |
| 17 | Claude Opus 4.8 | Anthropic | 159.0[157–162] | 91.0 | 80.0 | 98.3 | – | 53.0 | |
| 18 | Gemini 3.7 Flash | Google DeepMind | 158.0[156–161] | 94.8 | 71.6 | 97.2 | – | 69.2 | |
| 19 | Kimi K3Open weights | Moonshot | 157.9[155–161] | 93.1 | 72.2 | 97.2 | – | 50.6 | |
| 20 | Gemini 3.8 Flash | Google DeepMind | 157.4[155–161] | 95.4 | 68.4 | 98.9 | – | 69.7 | |
| 21 | Qwen 3.8 Max | Alibaba | 157.4[155–160] | 92.7 | 74.7 | 99.4 | – | 45.8 | |
| 22 | GPT-5.4 | OpenAI | 157.3[155–160] | 93.3 | 78.6 | 97.8 | 76.9 | 45.1 | |
| 23 | Muse Spark 1.3 | Meta AI | 157.2[155–160] | – | 74.4 | 99.2 | – | – | |
| 24 | GPT-5.3 Codex | OpenAI | 157.1[154–161] | – | – | – | 74.8 | – | |
| 25 | Claude Sonnet 5 | Anthropic | 156.9[154–160] | 90.5 | 65.6 | 94.7 | – | 33.7 | |
| 26 | Claude Opus 4.7 | Anthropic | 156.8[155–159] | 90.2 | 70.2 | 97.8 | 83.5 | 51.7 | |
| 27 | GPT-6 Luna | OpenAI | 156.8[154–159] | 90.5 | 78.9 | 98.9 | – | 41.4 | |
| 28 | Grok 4.6 | xAI | 156.6[155–159] | 94.0 | 66.0 | 99.2 | – | 49.3 | |
| 29 | GPT-5.6 Luna | OpenAI | 156.5[154–159] | 91.6 | 82.1 | 98.3 | – | 41.0 | |
| 30 | GPT-5.2 Pro | OpenAI | 156.3[154–160] | – | 74.0 | – | – | – | |
| 31 | GLM-5.3Open weights | Z.ai (Zhipu AI) | 156.0[154–158] | 90.9 | 68.8 | 91.1 | – | 41.0 | |
| 32 | DeepSeek V4.1 FlashOpen weights | DeepSeek | 155.9[154–159] | 89.8 | 67.4 | 98.3 | – | 42.9 | |
| 33 | Claude Opus 4.6 | Anthropic | 155.7[154–158] | 90.5 | 66.0 | 94.4 | 78.7 | 47.0 | |
| 34 | DeepSeek V4 Pro 0813Open weights | DeepSeek | 155.6[154–158] | 91.7 | 64.6 | 98.6 | – | 52.9 | |
| 35 | Gemini 3.1 Pro | Google DeepMind | 155.3[153–158] | 94.4 | 59.6 | 95.6 | 75.6 | 73.5 | |
| 36 | Muse Spark OpenOpen weights | Meta AI | 155.1[153–158] | 88.6 | 69.8 | 98.1 | – | 56.6 | |
| 37 | Qwen3.8 Max (0902) | Alibaba | 155.1[153–157] | 92.3 | 65.6 | 100.0 | – | 47.3 | |
| 38 | Muse Spark 1.2 | Meta AI | 155.1[153–158] | – | – | – | – | 60.3 | |
| 39 | Gemini 3.5 Flash | Google DeepMind | 154.9[153–157] | 92.8 | 62.8 | 95.6 | 79.3 | 66.2 | |
| 40 | DeepSeek V4 Flash 0731Open weights | DeepSeek | 154.8[152–157] | 91.0 | 57.5 | 94.4 | – | 33.6 | |
| 41 | Gemini 3.6 Flash | Google DeepMind | 154.5[153–157] | 94.1 | 58.9 | 94.2 | – | 66.2 | |
| 42 | Grok 4.7 | xAI | 154.4[153–157] | 92.7 | 53.0 | 98.1 | – | 56.0 | |
| 43 | Grok 4.5 | xAI | 154.1[152–156] | 93.4 | 57.2 | 97.8 | – | 48.3 | |
| 44 | Muse Spark 1.1 | Meta AI | 154.1[152–156] | – | – | – | – | 57.8 | |
| 45 | Qwen3.7-Max | Alibaba | 154.0[152–157] | 90.9 | 64.6 | 95.6 | 77.3 | 55.8 | |
| 46 | GPT-5.2 | OpenAI | 153.7[152–155] | 91.4 | 67.4 | 96.1 | 73.8 | 37.1 | |
| 47 | GLM-5.3-FlashOpen weights | Z.ai (Zhipu AI) | 153.1[151–156] | 90.2 | 55.8 | 93.9 | – | – | |
| 48 | Gemini 3 Pro | Google DeepMind | 153.1[151–155] | 92.6 | – | 91.4 | 72.9 | – | |
| 49 | Claude Sonnet 4.6 | Anthropic | 152.8[150–155] | 87.4 | – | 85.8 | 75.2 | 35.5 | |
| 50 | Grok 4.20 | xAI | 152.1[150–155] | 89.3 | 44.9 | 92.2 | – | 30.2 | |
| 51 | Muse Spark | Meta AI | 152.0[150–154] | 89.8 | – | 88.9 | – | – | |
| 52 | GLM-5.2Open weights | Z.ai (Zhipu AI) | 151.9[150–154] | 91.9 | 59.2 | 86.4 | 78.7 | 34.2 | |
| 53 | Gemini 3 Flash | Google DeepMind | 151.9[150–154] | 89.4 | 51.2 | 95.6 | 75.4 | 66.8 | |
| 54 | Qwen 3.8 27BOpen weights | Alibaba | 151.7[150–154] | – | – | – | – | – | |
| 55 | Kimi K2.6Open weights | Moonshot | 150.9[149–153] | 90.8 | 57.2 | 96.1 | 76.7 | 34.9 | |
| 56 | GPT-5 Pro | OpenAI | 150.6[149–154] | – | 55.8 | – | – | – | |
| 57 | Inkling-SmallOpen weights | Thinking Machines | 150.3[147–152] | 88.5 | 46.3 | 90.0 | – | 19.1 | |
| 58 | Kimi K2.7 CodeOpen weights | Moonshot | 150.1[148–152] | 87.9 | 54.0 | 95.6 | – | 36.5 | |
| 59 | Claude Opus 4.5 | Anthropic | 150.1[148–153] | 86.0 | 34.4 | 86.1 | 76.7 | 45.7 | |
| 60 | Mistral Large 4 | Mistral AI | 150.0[147–153] | 88.1 | 58.6 | 98.3 | – | 20.0 | |
| 61 | GPT-5 | OpenAI | 150.0 | 86.2 | 55.4 | 91.4 | 73.6 | 50.1 | |
| 62 | GLM-5.1Open weights | Z.ai (Zhipu AI) | 149.8[148–152] | 89.9 | 36.8 | 93.3 | 74.2 | 34.0 | |
| 63 | GPT-5.1 | OpenAI | 149.8[148–152] | 87.6 | – | 88.6 | 68.0 | 48.0 | |
| 64 | Qwen 3.6 Max (Preview) | Alibaba | 149.4[148–153] | 87.4 | – | 91.1 | 76.7 | 52.0 | |
| 65 | DeepSeek-V4-ProOpen weights | DeepSeek | 149.1[148–151] | 90.9 | 45.3 | 96.7 | 77.6 | 47.0 | |
| 66 | Grok 4.3 Beta | xAI | 148.8[147–150] | 88.8 | 42.8 | 93.3 | – | 33.2 | |
| 67 | GPT-5.4 Mini | OpenAI | 148.7[147–150] | 86.9 | 51.2 | 88.9 | – | 29.4 | |
| 68 | InklingOpen weights | Thinking Machines | 148.4[146–150] | 88.3 | 33.3 | 88.9 | – | 40.3 | |
| 69 | Kimi K2.5Open weights | Moonshot | 148.1[147–150] | 87.6 | – | 92.2 | 73.8 | 34.3 | |
| 70 | o3-pro | OpenAI | 147.3[146–150] | – | – | – | – | – | |
| 71 | Qwen 3.6 Plus | Alibaba | 147.3[145–149] | 88.4 | 38.2 | 93.3 | 57.9 | 44.1 | |
| 72 | Qwen3.7-Plus | Alibaba | 147.0[145–149] | 87.9 | 34.4 | 93.3 | – | – | |
| 73 | MiniMax-M3Open weights | MiniMax | 146.9[142–150] | 90.9 | – | 71.1 | – | – | |
| 74 | Qwen3.5 397B-A17BOpen weights | Alibaba | 146.8[145–149] | 86.4 | 31.2 | 88.9 | – | – | |
| 75 | Claude Sonnet 4.5 | Anthropic | 146.7[145–148] | 82.3 | 23.9 | 77.8 | 71.3 | 30.7 | |
| 76 | o3 | OpenAI | 146.7[145–149] | 81.8 | 33.3 | 84.4 | 62.3 | 49.4 | |
| 77 | MiniMax-M2.5Open weights | MiniMax | 146.6[142–148] | – | – | – | – | – | |
| 78 | Grok 4 | xAI | 146.2[144–148] | 87.0 | – | 84.0 | – | – | |
| 79 | Qwen3.6 27BOpen weights | Alibaba | 146.2[144–148] | 85.9 | 35.1 | 91.1 | – | – | |
| 80 | Qwen 3.5 Plus (hosted 397B-A17B) | Alibaba | 146.2[143–148] | 84.8 | – | 86.7 | – | 25.4 | |
| 81 | DeepSeek-V4-FlashOpen weights | DeepSeek | 146.1[144–148] | – | – | – | – | – | |
| 82 | DeepSeek-V3.2Open weights | DeepSeek | 146.1[144–147] | 83.4 | – | 87.8 | – | – | |
| 83 | GPT-5.4 Nano | OpenAI | 146.0[144–148] | 78.5 | 44.9 | 87.8 | – | 11.7 | |
| 84 | Nemotron 3 UltraOpen weights | Nvidia | 146.0[144–148] | 85.4 | – | 86.7 | – | – | |
| 85 | Kimi K2 ThinkingOpen weights | Moonshot | 145.8[143–148] | 84.2 | – | 83.1 | – | – | |
| 86 | GLM-5Open weights | Z.ai (Zhipu AI) | 145.8[144–148] | 87.8 | – | 80.0 | 72.1 | – | |
| 87 | MiniMax-M2.7Open weights | MiniMax | 145.8[138–148] | – | – | – | – | – | |
| 88 | o4-mini | OpenAI | 145.5[143–147] | 79.6 | 36.1 | 81.7 | – | 19.6 | |
| 89 | GPT-5 mini | OpenAI | 145.3[143–147] | 75.0 | 46.7 | 86.7 | 64.7 | 21.6 | |
| 90 | Gemini 2.5 Pro (Jun 2025) | Google DeepMind | 145.1[143–147] | 85.3 | 24.6 | 84.7 | 57.6 | – | |
| 91 | DeepSeek-V3.2-ExpOpen weights | DeepSeek | 144.9[142–147] | – | – | – | – | – | |
| 92 | Gemini 3.5 Flash-Lite | Google DeepMind | 144.9[142–146] | 83.3 | 26.0 | 71.1 | – | – | |
| 93 | Gemini 3.1 Flash-Lite | 144.3[142–146] | 81.8 | 27.7 | 80.0 | – | – | ||
| 94 | Grok 4 Fast | xAI | 144.1[142–146] | – | – | – | – | – | |
| 95 | Claude Opus 4.1 | Anthropic | 144.1[142–147] | 77.3 | 12.6 | 68.9 | 73.3 | – | |
| 96 | Qwen3.7 Flash | Alibaba | 144.1[142–146] | 82.3 | 19.3 | 86.7 | – | – | |
| 97 | Gemini 2.5 Pro (Mar 2025) | Google DeepMind | 144.0[142–147] | 83.8 | – | – | – | – | |
| 98 | Qwen3-235B-A22B-Thinking (Jul 2025)Open weights | Alibaba | 143.8[142–145] | 80.1 | – | 86.7 | – | 40.4 | |
| 99 | Qwen 3.6 35B-A3BOpen weights | Alibaba | 143.7[141–146] | 84.8 | 20.4 | 86.7 | – | – | |
| 100 | GLM-4.7Open weights | Z.ai (Zhipu AI) | 143.6[141–146] | 83.3 | – | 83.3 | – | 32.2 | |
| 101 | Qwen 3.5 Flash (hosted 35B-A3B) | Alibaba | 143.5[141–145] | 82.3 | 18.2 | 84.4 | – | 20.3 | |
| 102 | Gemma 4 31B ITOpen weights | Google DeepMind | 143.0[140–146] | 75.8 | – | 73.3 | – | 10.4 | |
| 103 | Gemini 2.5 Flash (Sep 2025) | Google DeepMind | 142.9[139–146] | – | – | – | – | – | |
| 104 | Qwen 3.6 Flash | Alibaba | 142.9[140–144] | 83.3 | 22.5 | 84.4 | – | 15.9 | |
| 105 | Claude Opus 4 | Anthropic | 142.5[140–144] | 76.3 | – | 64.4 | 70.7 | – | |
| 106 | Qwen3.5-35B-A3BOpen weights | Alibaba | 142.5[140–145] | 83.5 | – | 70.0 | – | – | |
| 107 | GPT-5.5 Instant | OpenAI | 142.4[140–145] | 82.5 | 26.3 | 68.1 | – | – | |
| 108 | Qwen3-Max | Alibaba | 142.3[140–145] | 72.6 | 18.9 | 73.3 | – | 48.7 | |
| 109 | Claude Haiku 4.5 | Anthropic | 142.2[139–144] | 71.2 | – | 66.7 | – | 13.2 | |
| 110 | Gemini 2.5 Pro (May 2025) | Google DeepMind | 142.2[138–145] | 66.7 | – | – | – | – | |
| 111 | Gemma 4 26B A4BOpen weights | Google DeepMind | 141.9[139–143] | 73.2 | – | 82.2 | – | – | |
| 112 | o1 | OpenAI | 141.7[140–143] | 76.8 | 14.7 | 73.3 | – | 41.1 | |
| 113 | Claude Sonnet 4 | Anthropic | 141.6[139–143] | 79.2 | – | 71.1 | – | – | |
| 114 | Gemini 2.5 Flash (May 2025) | Google DeepMind | 141.4[139–143] | – | – | 70.8 | – | – | |
| 115 | Mistral Medium 3.5Open weights | Mistral AI | 141.4[139–144] | – | – | – | – | – | |
| 116 | DeepSeek-R1 (May 2025)Open weights | DeepSeek | 141.1[139–143] | 76.3 | – | 66.4 | – | – | |
| 117 | Claude 3.7 Sonnet | Anthropic | 141.0[139–143] | 79.7 | – | 57.8 | 61.0 | – | |
| 118 | Gemini 2.5 Flash (Jun 2025) | Google DeepMind | 140.6[138–142] | – | – | – | – | – | |
| 119 | o3-mini | OpenAI | 140.2[137–142] | 77.0 | 18.6 | 76.9 | – | 15.3 | |
| 120 | Grok-3 mini | xAI | 140.2[138–142] | 76.3 | – | 77.8 | – | – | |
| 121 | Kimi K2 (Jul 2025)Open weights | Moonshot | 140.0[136–142] | – | – | – | – | – | |
| 122 | Gemini 2.5 Flash (Apr 2025) | Google DeepMind | 139.8[136–142] | – | – | 73.1 | – | – | |
| 123 | gpt-oss-120bOpen weights | OpenAI | 139.8[136–142] | 75.8 | – | 88.9 | – | – | |
| 124 | DeepSeek-V3.1Open weights | DeepSeek | 139.8[136–143] | – | – | – | – | – | |
| 125 | Qwen3-30B-A3B-Thinking (Jul 2025)Open weights | Alibaba | 139.6[135–141] | 70.1 | – | 70.3 | – | – | |
| 126 | Qwen3.5-9BOpen weights | Alibaba | 139.3[136–141] | 79.0 | – | 61.7 | – | – | |
| 127 | GPT-5 nano | OpenAI | 139.2[134–141] | 69.4 | 20.0 | 81.1 | – | 11.7 | |
| 128 | Qwen3-235B-A22BOpen weights | Alibaba | 139.2[135–141] | 70.7 | – | – | – | – | |
| 129 | DeepSeek-R1Open weights | DeepSeek | 138.8[136–140] | 71.7 | – | 53.3 | – | – | |
| 130 | Qwen3-235B-A22B-Instruct (Jul 2025)Open weights | Alibaba | 138.8[135–141] | – | – | – | – | – | |
| 131 | Qwen3-32BOpen weights | Alibaba | 138.4[135–140] | 65.7 | – | 66.9 | – | – | |
| 132 | Grok 3 | xAI | 138.2[135–140] | 75.8 | – | 55.6 | – | – | |
| 133 | Qwen3-14BOpen weights | Alibaba | 138.2[133–140] | 63.8 | – | 66.4 | – | – | |
| 134 | gpt-oss-20bOpen weights | OpenAI | 137.7[132–140] | 60.8 | – | 65.3 | – | – | |
| 135 | QwQ-32BOpen weights | Alibaba | 137.5[133–142] | 65.3 | – | 59.2 | – | – | |
| 136 | Qwen3-30B-A3B-Instruct (Jul 2025)Open weights | Alibaba | 137.3[132–140] | 55.6 | – | 62.2 | – | – | |
| 137 | GPT-4.5 | OpenAI | 136.7[134–139] | 68.7 | – | 37.8 | – | – | |
| 138 | GPT-4.1 | OpenAI | 136.7[134–138] | 66.9 | 6.0 | 38.3 | 48.5 | 31.1 | |
| 139 | Qwen3-30B-A3BOpen weights | Alibaba | 136.1[131–139] | 61.7 | – | 62.8 | – | – | |
| 140 | Qwen3-8BOpen weights | Alibaba | 136.1[130–138] | 56.8 | – | 56.1 | – | – | |
| 141 | DeepSeek-V3 (Mar 2025)Open weights | DeepSeek | 135.8[132–138] | 67.6 | – | 37.8 | – | – | |
| 142 | o1-mini | OpenAI | 135.7[132–137] | 62.4 | – | 46.9 | – | – | |
| 143 | Gemini 2.0 Flash Thinking (Jan 2025) | Google DeepMind,Google | 135.2[131–138] | 57.1 | – | 57.8 | – | – | |
| 144 | Gemini 2.0 Pro | Google DeepMind | 134.9[131–137] | 65.7 | – | – | – | – | |
| 145 | GPT-4.1 mini | OpenAI | 134.9[131–136] | 65.8 | 6.7 | 44.7 | – | 12.7 | |
| 146 | Gemini 2.0 Flash (Feb 2025) | Google DeepMind,Google | 134.6[131–136] | 64.1 | – | 31.1 | – | – | |
| 147 | Gemini 2.0 Flash (Dec 2024) | Google DeepMind,Google | 134.6[124–136] | – | – | – | – | – | |
| 148 | o1-preview | OpenAI | 134.6[131–138] | 50.3 | – | 31.1 | – | – | |
| 149 | Mistral Medium 3 | Mistral AI | 134.0[130–135] | 59.5 | – | 32.2 | – | – | |
| 150 | Gemini 2.5 Flash-Lite (Jun 2025) | Google DeepMind | 133.8[130–136] | – | – | – | – | – | |
| 151 | Claude 3.5 Sonnet (October 2024) | Anthropic | 133.4[129–138] | 55.3 | – | 8.5 | – | – | |
| 152 | Magistral Small 1.0Open weights | Mistral AI | 133.1[129–134] | 56.1 | – | 30.0 | – | – | |
| 153 | Qwen2.5-Max | Alibaba | 132.5[129–136] | 56.1 | – | 16.1 | – | – | |
| 154 | DeepSeek-V3Open weights | DeepSeek | 132.3[128–135] | 56.5 | – | 15.8 | – | – | |
| 155 | Llama 4 MaverickOpen weights | Meta AI | 132.2[128–134] | 67.0 | – | 20.6 | – | – | |
| 156 | Gemini 1.5 Pro (Sept 2024) | Google DeepMind | 131.7[128–133] | 57.2 | – | 23.1 | – | – | |
| 157 | Phi-4Open weights | Microsoft Research | 130.5[125–132] | 56.1 | – | 13.8 | – | – | |
| 158 | Grok-2 (Dec 2024) | xAI | 130.5[126–132] | 53.8 | – | 11.5 | – | – | |
| 159 | Claude 3.5 Sonnet | Anthropic | 130.0 | 54.0 | – | 6.5 | – | – | |
| 160 | Gemma 3 27BOpen weights | Google DeepMind | 130.0[124–132] | 47.7 | – | 22.5 | – | – | |
| 161 | Llama 4 ScoutOpen weights | Meta AI | 129.6[125–131] | 51.8 | – | 7.8 | – | – | |
| 162 | GPT-4.1 nano | OpenAI | 129.5[123–132] | 48.9 | – | 28.9 | – | 6.0 | |
| 163 | Gemini 1.5 Flash (Sep 2024) | Google DeepMind | 129.3[124–131] | 47.3 | – | 16.3 | – | – | |
| 164 | GPT-4o (May 2024) | OpenAI | 129.0[125–131] | 48.9 | – | 6.3 | – | – | |
| 165 | Qwen2.5-72BOpen weights | Alibaba | 129.0[123–131] | 49.1 | – | 8.1 | – | – | |
| 166 | GPT-4o (Aug 2024) | OpenAI | 128.8[124–131] | 49.2 | 0.4 | 6.4 | – | 26.0 | |
| 167 | Llama 3.1-405BOpen weights | Meta AI | 128.8[124–131] | 50.9 | – | 9.7 | – | – | |
| 168 | GPT-4o (Nov 2024) | OpenAI | 128.7[124–131] | 47.9 | – | 6.3 | 31.0 | – | |
| 169 | Mistral Large 2 (Nov 2024)Open weights | Mistral AI | 128.6[124–131] | 51.3 | – | 7.8 | – | – | |
| 170 | Mistral Large 2 (Jul 2024)Open weights | Mistral AI | 127.6[122–130] | 49.0 | – | 8.5 | – | – | |
| 171 | Mistral Small 3.1Open weights | Mistral AI | 127.5[122–129] | 47.5 | – | 5.8 | – | – | |
| 172 | Llama 3.3 70BOpen weights | Meta AI | 127.3[123–129] | 47.4 | – | 5.1 | – | – | |
| 173 | GPT-4 Turbo (Apr 2024) | OpenAI | 127.3[122–129] | 46.6 | 0.7 | 6.7 | – | – | |
| 174 | Claude 3.5 Haiku | Anthropic | 127.1[121–130] | 38.1 | – | 4.3 | – | – | |
| 175 | Gemini 1.5 Pro (May 2024) | Google DeepMind | 126.9[122–129] | 45.9 | – | 6.8 | – | – | |
| 176 | Claude 3 Opus | Anthropic | 126.9[122–130] | 47.2 | – | 4.7 | – | 12.6 | |
| 177 | GPT-4o mini | OpenAI | 126.5[120–128] | 37.7 | 0.7 | 6.9 | – | 8.3 | |
| 178 | GPT-4 Turbo (Nov 2023) | OpenAI | 126.4[121–128] | 42.4 | – | – | – | – | |
| 179 | Llama 3.1-70BOpen weights | Meta AI | 125.9[121–128] | 44.2 | – | 3.6 | – | – | |
| 180 | GPT-4 (Mar 2023) | OpenAI | 125.7[121–131] | 35.7 | – | 0.6 | – | – | |
| 181 | Llama 3.2 90BOpen weights | Meta AI | 125.5[120–127] | 41.0 | – | 2.6 | – | – | |
| 182 | Qwen2-72BOpen weights | Alibaba | 125.3[119–127] | 40.8 | – | – | – | – | |
| 183 | DeepSeek-V2 (MoE-236B, May 2024)Open weights | DeepSeek | 124.8[119–128] | – | – | – | – | – | |
| 184 | Gemma 3 12BOpen weights | Google DeepMind | 123.5[114–129] | 39.5 | – | 16.7 | – | – | |
| 185 | GPT-4 (Jun 2023) | OpenAI | 123.1[117–127] | 30.7 | – | 1.1 | – | – | |
| 186 | Llama 3-70BOpen weights | Meta AI | 123.0[117–126] | 40.6 | – | 4.3 | – | – | |
| 187 | Gemini 1.5 Flash (May 2024) | Google DeepMind | 122.7[116–125] | 40.4 | – | 3.9 | – | – | |
| 188 | Mistral Large | Mistral AI | 122.1[115–125] | 38.8 | – | 1.9 | – | – | |
| 189 | Gemma 2 27BOpen weights | Google DeepMind | 122.1[116–124] | 36.5 | – | 1.4 | – | – | |
| 190 | Mixtral 8x22BOpen weights | Mistral AI | 122.0[115–124] | 34.1 | – | – | – | – | |
| 191 | phi-3-small 7.4BOpen weights | Microsoft | 121.8[114–125] | – | – | – | – | – | |
| 192 | phi-3-medium 14BOpen weights | Microsoft | 121.2[114–124] | 27.6 | – | – | – | – | |
| 193 | Claude 3 Sonnet | Anthropic | 120.8[113–124] | 40.6 | – | 2.5 | – | – | |
| 194 | Claude Instant | Anthropic | 120.2[113–123] | – | – | – | – | – | |
| 195 | Claude 2 | Anthropic | 120.2[114–124] | 34.7 | – | 2.5 | – | – | |
| 196 | Gemma 2 9BOpen weights | Google DeepMind | 119.8[112–122] | 27.5 | – | 0.6 | – | – | |
| 197 | Qwen2.5-Coder-32BOpen weights | Alibaba | 119.5[111–125] | – | – | – | – | – | |
| 198 | Claude 2.1 | Anthropic | 119.3[112–122] | 33.0 | – | 1.9 | – | – | |
| 199 | Command R+Open weights | Cohere,Cohere Labs (formerly Cohere for AI) | 119.2[110–125] | – | – | – | – | – | |
| 200 | Claude 3 Haiku | Anthropic | 118.8[111–122] | 36.3 | – | 1.8 | – | – | |
| 201 | Mistral NeMoOpen weights | Mistral AI | 118.7[111–121] | 29.9 | – | – | – | – | |
| 202 | Qwen2.5-7BOpen weights | Alibaba | 118.5[110–122] | 35.5 | – | 2.5 | – | – | |
| 203 | GPT-3.5 Turbo (Nov 2023) | OpenAI | 118.5[111–121] | 28.0 | – | – | – | – | |
| 204 | Mixtral 8x7BOpen weights | Mistral AI | 118.4[111–121] | 30.6 | – | – | – | – | |
| 205 | Ministral 3B | Mistral AI | 118.0[106–122] | 25.3 | – | – | – | – | |
| 206 | Yi-34BOpen weights | 01.AI | 117.3[109–121] | 14.7 | – | – | – | – | |
| 207 | phi-3-mini 3.8BOpen weights | Microsoft | 117.3[110–121] | – | – | – | – | – | |
| 208 | Gemini 1.0 Pro | Google DeepMind | 117.1[109–120] | 34.0 | – | 1.1 | – | – | |
| 209 | Stable Beluga 2Open weights | Stability AI | 117.0[110–120] | – | – | – | – | – | |
| 210 | Llama 3.1-8BOpen weights | Meta AI | 116.5[106–122] | 27.0 | – | 1.7 | – | – | |
| 211 | Llama 3-8BOpen weights | Meta AI | 116.4[108–119] | 26.1 | – | 1.9 | – | – | |
| 212 | Qwen2.5-Coder-14B | Unknown | 116.3[106–122] | – | – | – | – | – | |
| 213 | Gemma 3 4BOpen weights | Google DeepMind | 115.9[97–124] | 23.2 | – | 7.5 | – | – | |
| 214 | GPT-3.5 Turbo (Jan 2024) | OpenAI | 115.7[108–119] | 27.2 | 0.0 | 2.2 | – | – | |
| 215 | PaLM 2-L | Unknown | 114.9[107–128] | – | – | – | – | – | |
| 216 | Llama 2-70BOpen weights | Meta AI | 113.7[105–117] | 26.3 | – | 0.0 | – | – | |
| 217 | GPT-3.5 Turbo (Jun 2023) | OpenAI | 113.2[105–118] | – | – | – | – | – | |
| 218 | Qwen2.5-Coder (7B)Open weights | Alibaba | 113.0[103–119] | – | – | – | – | – | |
| 219 | Qwen-14BOpen weights | Alibaba | 112.9[104–117] | – | – | – | – | – | |
| 220 | Mistral 7B v0.1Open weights | Mistral AI | 112.0[103–116] | – | – | – | – | – | |
| 221 | Falcon-180BOpen weights | Technology Innovation Institute | 112.0[102–120] | – | – | – | – | – | |
| 222 | internlm-20b | Unknown | 111.9[103–117] | – | – | – | – | – | |
| 223 | Gemma 7BOpen weights | Google DeepMind | 111.8[103–117] | – | – | – | – | – | |
| 224 | LLaMA-65BOpen weights | Meta AI | 110.0[101–114] | – | – | – | – | – | |
| 225 | Falcon 2 11BOpen weights | Technology Innovation Institute | 109.4[101–116] | – | – | – | – | – | |
| 226 | Mistral 7B v0.3Open weights | Mistral AI | 108.9[98–113] | 15.2 | – | 0.3 | – | – | |
| 227 | DeepSeek-Coder-V2-Lite-Base | Unknown | 108.7[98–114] | – | – | – | – | – | |
| 228 | PaLM 2-M | Unknown | 108.1[100–117] | – | – | – | – | – | |
| 229 | Phi-2Open weights | Microsoft | 107.7[86–114] | – | – | – | – | – | |
| 230 | Qwen2.5-Coder-3B | Unknown | 107.5[93–115] | – | – | – | – | – | |
| 231 | Nemotron-4 15B | Nvidia | 107.5[98–113] | – | – | – | – | – | |
| 232 | Yi-9B | Unknown | 107.4[98–115] | – | – | – | – | – | |
| 233 | LLaMA-33BOpen weights | Meta AI | 107.2[97–112] | – | – | – | – | – | |
| 234 | Qwen-7BOpen weights | Alibaba | 106.6[95–112] | – | – | – | – | – | |
| 235 | PaLM 2-S | Unknown | 105.9[95–114] | – | – | – | – | – | |
| 236 | Llama 2-13BOpen weights | Meta AI | 105.6[95–111] | – | – | – | – | – | |
| 237 | Llama 2-34B | Meta AI | 104.9[94–110] | – | – | – | – | – | |
| 238 | StarCoder 2 15BOpen weights | Hugging Face,ServiceNow,Nvidia,BigCode | 104.7[91–112] | – | – | – | – | – | |
| 239 | Yi 6BOpen weights | 01.AI | 104.5[94–110] | – | – | – | – | – | |
| 240 | Falcon-40BOpen weights | Technology Innovation Institute | 104.2[94–110] | – | – | – | – | – | |
| 241 | Baichuan2-13BOpen weights | Baichuan | 102.9[92–110] | – | – | – | – | – | |
| 242 | Qwen2.5-Coder (1.5B)Open weights | Alibaba | 102.7[85–110] | – | – | – | – | – | |
| 243 | internlm-7b | Unknown | 102.6[91–109] | – | – | – | – | – | |
| 244 | INTELLECT-1 | Prime Intellect,Hugging Face,Arcee AI | 100.5[89–106] | – | – | – | – | – | |
| 245 | MPT-30BOpen weights | MosaicML | 100.3[89–106] | – | – | – | – | – | |
| 246 | LLaMA-13BOpen weights | Meta AI | 100.3[89–106] | – | – | – | – | – | |
| 247 | Llama 2-7BOpen weights | Meta AI | 98.6[87–106] | – | – | – | – | – | |
| 248 | chatglm2-6b | Unknown | 98.6[79–105] | – | – | – | – | – | |
| 249 | LLaMA-7BOpen weights | Meta AI | 96.6[84–104] | – | – | – | – | – | |
| 250 | Baichuan 2-7BOpen weights | Baichuan | 95.9[83–105] | – | – | – | – | – | |
| 251 | DeepSeek Coder 33BOpen weights | DeepSeek,Peking University | 95.9[83–103] | – | – | – | – | – | |
| 252 | Falcon-7BOpen weights | Technology Innovation Institute | 94.7[82–102] | – | – | – | – | – | |
| 253 | CodeQwen1.5-7B | Unknown | 94.5[79–103] | – | – | – | – | – | |
| 254 | vicuna-13b-v1.1 | Unknown | 94.3[82–102] | – | – | – | – | – | |
| 255 | MPT-7BOpen weights | MosaicML | 94.2[80–101] | – | – | – | – | – | |
| 256 | Gemma 2BOpen weights | Google DeepMind | 93.8[81–100] | – | – | – | – | – | |
| 257 | StarCoder 2 7BOpen weights | Hugging Face,ServiceNow,Nvidia,BigCode | 93.1[76–102] | – | – | – | – | – | |
| 258 | XGen-7BOpen weights | Salesforce | 93.0[79–100] | – | – | – | – | – | |
| 259 | Qwen-1_8B | Unknown | 92.5[71–101] | – | – | – | – | – | |
| 260 | open_llama_7b | Unknown | 91.2[78–98] | – | – | – | – | – | |
| 261 | Phi-1.5Open weights | Microsoft | 91.1[65–100] | – | – | – | – | – | |
| 262 | Baichuan1-7BOpen weights | Baichuan | 90.0[76–98] | – | – | – | – | – | |
| 263 | RedPajama-INCITE-7B-Base | Unknown | 90.0[75–98] | – | – | – | – | – | |
| 264 | Dolly 2.0-12bOpen weights | Databricks | 89.2[72–98] | – | – | – | – | – | |
| 265 | DeepSeek Coder 6.7BOpen weights | DeepSeek,Peking University | 89.1[72–98] | – | – | – | – | – | |
| 266 | StarCoder 2 3BOpen weights | Hugging Face,ServiceNow,Nvidia,BigCode | 88.3[72–97] | – | – | – | – | – | |
| 267 | Qwen2.5-Coder-0.5B | Unknown | 87.6[61–100] | – | – | – | – | – | |
| 268 | Cerebras-GPT-13BOpen weights | Cerebras Systems | 82.7[66–91] | – | – | – | – | – | |
| 269 | DeepSeek Coder 1.3BOpen weights | DeepSeek,Peking University | 62.6[42–76] | – | – | – | – | – | |
| 270 | stablelm-tuned-alpha-7b | Unknown | 54.8[29–71] | – | – | – | – | – |
No models match.
Source: Epoch AI Benchmarking Hub · CC BY 4.0 · Last updated · All sources
Citation: Epoch AI, ‘Capabilities & benchmarking’. Published online at epoch.ai. Retrieved from https://epoch.ai/benchmarks. We show Epoch-run results only; benchmarks Epoch collects from other projects keep their own licenses and aren't included.