MATH (Hendrycks et al.) — Heterogeneous bucket (222 models); max=100 suggests mixed conventions. Not merged w/ MATH variants.
| # | Model | Vendor | Best score | Runs | Last seen |
|---|---|---|---|---|---|
| 1 | DeepSeek-R1 | DeepSeek | 97.3 | 1 | 2026-05-01 |
| 2 | o1 | OpenAI | 96.4 | 1 | 2026-05-01 |
| 3 | Gemini 1.5 Pro | 92 | 3 | 2026-06-06 | |
| 4 | Kimi VL A3B Thinking | Moonshot | 91.8 | 1 | 2026-06-05 |
| 5 | DeepSeek-V3 | DeepSeek | 91.2 | 4 | 2026-06-06 |
| 6 | Gemini 1.5 Flash (001) | 90.8 | 2 | 2026-05-10 | |
| 7 | Gemini 2.0 Flash Exp | 90.1 | 2 | 2026-06-06 | |
| 8 | Gemini 2.0 Flash (experimental) | 90.1 | 1 | 2026-06-24 | |
| 9 | OpenAI o1-mini | OpenAI | 90 | 1 | 2026-05-01 |
| 10 | MAI-Thinking-1 | Microsoft | 89.5 | 1 | 2026-08-14 |
| 11 | Gemma 3 27B | 89 | 1 | 2026-06-05 | |
| 12 | Qwen2.5 Instruct 72B | Alibaba | 88.4 | 4 | 2026-06-06 |
| 13 | Claude 3.5 Haiku | Anthropic | 87.2 | 2 | 2026-06-19 |
| 14 | MiMo V2.5 Pro Base | Xiaomi | 86.2 | 3 | 2026-06-05 |
| 15 | GPT4 | OpenAI | 85.7 | 3 | 2026-08-12 |
| 16 | GPT-4 Turbo | OpenAI | 85.7 | 2 | 2026-06-18 |
| 17 | GPT-4o | OpenAI | 85.3 | 5 | 2026-06-06 |
| 18 | Qwen2.5 Instruct Turbo (7B) | Alibaba | 83.5 | 1 | 2026-05-10 |
| 19 | Llama 3.1 Instruct 405B | Meta | 82.7 | 6 | 2026-06-06 |
| 20 | Qwen2.5 VL 32B | Alibaba | 82.2 | 1 | 2026-06-05 |
| 21 | Nova Pro (Non-Reasoning) | Amazon | 82.1 | 2 | 2026-06-19 |
| 22 | Nemotron 3 Ultra 550B A55B Base | NVIDIA | 82 | 1 | 2026-06-05 |
| 23 | NVIDIA-Nemotron-3-Ultra-550B-A55B-Base-BF16 | NVIDIA | 82 | 1 | 2026-08-24 |
| 24 | Claude 3.5 Sonnet | Anthropic | 81.3 | 3 | 2026-06-06 |
| 25 | Llama 3.3 Instruct Turbo (70B) | Meta | 80.8 | 1 | 2026-05-10 |
| 26 | Phi 4 | Microsoft | 80.4 | 1 | 2026-05-30 |
| 27 | GPT-4o mini | OpenAI | 80.2 | 2 | 2026-05-30 |
| 28 | Qwen3 1.7B | Alibaba | 79.3 | 1 | 2026-05-19 |
| 29 | Llama 3.2 Vision Instruct Turbo (90B) | Meta | 79.1 | 1 | 2026-05-10 |
| 30 | Qwen2 Instruct (72B) | Alibaba | 79 | 1 | 2026-05-10 |
| 31 | Llama 3.1 Instruct Turbo (70B) | Meta | 78.3 | 1 | 2026-05-10 |
| 32 | Amazon Nova Lite | Amazon | 77.9 | 1 | 2026-06-23 |
| 33 | MiniMax Text 01 | MiniMax | 77.4 | 1 | 2026-06-06 |
| 34 | Hunyuan Large Inst. | Tencent | 77.4 | 1 | 2026-05-31 |
| 35 | Llama 3.3 70B Instruct | Meta | 77 | 2 | 2026-05-30 |
| 36 | Palmyra-X-004 | Writer | 76.7 | 1 | 2026-05-10 |
| 37 | Hy3 preview-Base | Tencent | 76.3 | 2 | 2026-06-05 |
| 38 | Claude 3 Opus | Anthropic | 76 | 1 | 2026-05-10 |
| 39 | Nova Micro (Non-Reasoning) | Amazon | 76 | 2 | 2026-06-19 |
| 40 | Qwen 2.5 14B | Alibaba | 75.6 | 2 | 2026-05-30 |
| 41 | Mistral Large | Mistral | 75 | 1 | 2026-05-10 |
| 42 | DeepSeek-V2.5 | DeepSeek | 74.7 | 1 | 2026-05-31 |
| 43 | Gemma 2 Instruct (27B) | 74.6 | 1 | 2026-05-10 | |
| 44 | Llama 3.2 Vision Instruct Turbo (11B) | Meta | 73.9 | 1 | 2026-05-10 |
| 45 | Gemma 2 Instruct (9B) | 72.4 | 1 | 2026-05-10 | |
| 46 | Palmyra X V3 (72B) | Writer | 72.3 | 1 | 2026-05-10 |
| 47 | Kimi K2 Base | Moonshot | 71.2 | 10 | 2026-06-13 |
| 48 | Yi Large (Preview) | 01.AI | 71.2 | 1 | 2026-05-10 |
| 49 | MiMo V2 Flash Base | Xiaomi | 71 | 4 | 2026-06-13 |
| 50 | Phi 3 Small 8K Instruct | Microsoft | 70.3 | 1 | 2026-05-10 |
| 51 | Llama 3.1 Instruct Turbo (8B) | Meta | 70.3 | 1 | 2026-05-10 |
| 52 | Hunyuan Large | Tencent | 69.8 | 1 | 2026-05-31 |
| 53 | Jamba 1.5 Large | AI21 | 69.2 | 1 | 2026-05-10 |
| 54 | Qwen2.5 Math 1.5B | Alibaba | 68 | 1 | 2026-05-06 |
| 55 | Llama 3.1 70B Instruct | Meta | 68 | 2 | 2026-05-31 |
| 56 | MiMo V2.5 Base | Xiaomi | 67.7 | 3 | 2026-06-05 |
| 57 | Mistral Large 2 | Mistral | 67.7 | 1 | 2026-05-10 |
| 58 | Mistral Nemo | Mistral | 66.8 | 1 | 2026-05-10 |
| 59 | Step3.5 Flash Base | StepFun | 66.8 | 2 | 2026-06-05 |
| 60 | Llama 3 (70B) | Meta | 66.3 | 1 | 2026-05-10 |
| 61 | DeepSeek-V4-Pro-Base | DeepSeek | 64.5 | 3 | 2026-06-05 |
| 62 | Mistral Large 3 675B Base 2512 | Mistral | 62.9 | 1 | 2026-06-05 |
| 63 | DeepSeek-V3.1-Base | DeepSeek | 62.6 | 4 | 2026-06-13 |
| 64 | GLM-4.5-Base | Z.ai | 62.6 | 5 | 2026-06-05 |
| 65 | DeepSeek-V3.2-Exp-Base | DeepSeek | 62.5 | 5 | 2026-06-13 |
| 66 | Mistral Small (2402) | Mistral | 62 | 1 | 2026-05-10 |
| 67 | DeepSeek-Chat | DeepSeek | 61.5 | 1 | 2026-05-10 |
| 68 | Phi 3 Medium 4K Instruct | Microsoft | 61.1 | 2 | 2026-05-30 |
| 69 | DeepSeek-V3-Base | DeepSeek | 59.4 | 2 | 2026-06-05 |
| 70 | Palmyra X V2 (33B) | Writer | 58 | 1 | 2026-05-10 |
| 71 | DeepSeek-V4-Flash-Base | DeepSeek | 57.4 | 3 | 2026-06-05 |
| 72 | Solar Pro | Upstage | 56.7 | 1 | 2026-05-10 |
| 73 | Mistral Medium (Non-Reasoning) | Mistral | 56.5 | 1 | 2026-06-23 |
| 74 | Arctic Instruct | Snowflake | 51.9 | 1 | 2026-05-10 |
| 75 | Llama 3.1 8B Instruct | Meta | 51.9 | 1 | 2026-05-01 |
| 76 | Gemma (7B) | 50 | 1 | 2026-05-10 | |
| 77 | Moonlight | Moonshot | 45.3 | 2 | 2026-06-12 |
| 78 | DeepSeek-V2 | DeepSeek | 43.6 | 2 | 2026-05-31 |
| 79 | Qwen2.5 3B | Alibaba | 42.6 | 2 | 2026-06-05 |
| 80 | Command R+ (Apr '24) | Cohere | 40.3 | 1 | 2026-05-10 |
| 81 | Llama 3 Instruct 8B | Meta | 39.1 | 1 | 2026-05-10 |
| 82 | Jamba Instruct | AI21 | 38 | 1 | 2026-05-10 |
| 83 | DBRX Instruct | Databricks | 35.8 | 1 | 2026-05-10 |
| 84 | Jamba 1.5 Mini | AI21 | 31.8 | 1 | 2026-05-10 |
| 85 | Mistral V0.1 (7B) | Mistral | 29.7 | 1 | 2026-05-10 |
| 86 | Mistral Instruct V0.3 (7B) | Mistral | 28.9 | 1 | 2026-05-10 |
| 87 | Command R | Cohere | 26.6 | 1 | 2026-05-10 |
| 88 | Llama (65B) | Meta | 25.7 | 1 | 2026-05-10 |
| 89 | Phi 2 | Microsoft | 25.5 | 1 | 2026-05-10 |
| 90 | Grok | SpaceXAI | 23.9 | 1 | 2026-08-12 |
| 91 | Command | Cohere | 23.6 | 1 | 2026-05-10 |
| 92 | Inflection-1 | SpaceXAI | 16 | 1 | 2026-08-12 |
| 93 | Claude 3 Haiku | Anthropic | 13.1 | 1 | 2026-05-10 |
| 94 | Jurassic 2 Jumbo (178B) | AI21 | 10.3 | 1 | 2026-05-10 |
| 95 | Command Light | Cohere | 9.8 | 1 | 2026-05-10 |
| 96 | Llama 3.2 3B Instruct | Meta | 8.5 | 2 | 2026-06-05 |
| 97 | Claude 3 Sonnet | Anthropic | 8.4 | 1 | 2026-05-10 |
| 98 | Luminous Supreme (70B) | Aleph Alpha | 7.8 | 1 | 2026-05-10 |
| 99 | Jurassic 2 Grande (17B) | AI21 | 6.4 | 1 | 2026-05-10 |
| 100 | Luminous Extended (30B) | Aleph Alpha | 4 | 1 | 2026-05-10 |
Best tracked score per model (default configuration; source-attributed and verification-tiered). Open a model for its full benchmark surface, provenance and pricing.