MATH Level 5
Hardest level of the MATH competition dataset. Saturated, so it gets a low weight.
| Category | Math |
|---|---|
| Metric | Accuracy |
| Source | Epoch AI (evaluator) |
| License | CC BY 4.0 |
| Benchmark page | epoch.ai |
| Status | Scored |
| Weight | 0.5 within Math |
| Fit | Discrimination 2.00, noise 0.23 |
Results
| # | Model | Provider | Result | Date |
|---|---|---|---|---|
| 1 | GPT-5 High | OpenAI | 98.1% | 2025-10-29 |
| 2 | GPT-5 Medium | OpenAI | 97.9% | 2025-08-20 |
| 3 | GPT-5 Mini High | OpenAI | 97.9% | 2025-10-30 |
| 4 | o4-mini High | OpenAI | 97.8% | 2025-04-16 |
| 5 | o3 High | OpenAI | 97.8% | 2025-04-16 |
| 6 | Claude Sonnet 4.5 Thinking | Anthropic | 97.7% | 2025-10-21 |
| 7 | Qwen3 Max | Alibaba | 97.1% | 2025-10-09 |
| 8 | GPT-5 Mini Medium | OpenAI | 96.8% | 2025-08-20 |
| 9 | DeepSeek R1 0528 | DeepSeek | 96.6% | 2025-05-29 |
| 10 | o3-mini High | OpenAI | 96.5% | 2025-02-13 |
| 11 | Claude Haiku 4.5 Thinking | Anthropic | 96.4% | 2025-10-22 |
| 12 | Gemini 2.5 Pro Preview 06-05 | 95.9% | 2025-05-08 | |
| 13 | GPT-5 Nano Medium | OpenAI | 95.2% | 2025-08-20 |
| 14 | o3-mini Medium | OpenAI | 95.2% | 2025-01-31 |
| 15 | GPT-5 Nano High | OpenAI | 94.9% | 2025-08-20 |
| 16 | o1 High | OpenAI | 94.7% | 2025-02-13 |
| 17 | o1 Medium | OpenAI | 94.4% | 2025-01-27 |
| 18 | DeepSeek-R1 | DeepSeek | 93.1% | 2025-01-31 |
| 19 | Claude Sonnet 3.7 Thinking | Anthropic | 91.2% | 2025-03-13 |
| 20 | Grok-3 mini Low | xAI | 90.9% | 2025-04-10 |
| 21 | DeepSeek-R1-Distill-Llama-70B | DeepSeek | 89.9% | 2025-03-10 |
| 22 | o1-mini High | OpenAI | 89.2% | 2025-02-13 |
| 23 | Grok 3 | xAI | 88.8% | 2025-04-10 |
| 24 | Grok-3 mini High | xAI | 88.1% | 2025-04-10 |
| 25 | GPT-4.1 mini | OpenAI | 87.3% | 2025-04-14 |
| 26 | DeepSeek-R1-Distill-Qwen-14B | DeepSeek | 87.1% | 2025-03-11 |
| 27 | Claude Haiku 4.5 | Anthropic | 86.9% | 2025-10-16 |
| 28 | Claude Opus 4 | Anthropic | 85.1% | 2025-05-22 |
| 29 | Claude Sonnet 4 | Anthropic | 84.4% | 2025-05-22 |
| 30 | o1-mini Medium | OpenAI | 84.3% | 2025-01-27 |
| 31 | Gemini 2.0 Pro | 83.5% | 2025-02-06 | |
| 32 | GPT-4.1 | OpenAI | 83.0% | 2025-04-14 |
| 33 | Gemini 2.0 Flash | 82.2% | 2025-02-06 | |
| 34 | o1-preview | OpenAI | 81.7% | 2025-01-27 |
| 35 | Mistral Medium 3 | Mistral | 81.6% | 2025-05-07 |
| 36 | GPT-4.5 | OpenAI | 78.6% | 2025-02-28 |
| 37 | DeepSeek V3 0324 | DeepSeek | 75.6% | 2025-04-01 |
| 38 | Gemma 3 27B IT | 74.0% | 2025-03-13 | |
| 39 | Llama 4 Maverick | Meta | 73.0% | 2025-04-08 |
| 40 | Gemini 1.5 Pro (Sept 2024) | 70.4% | 2025-01-27 | |
| 41 | GPT-4.1 nano | OpenAI | 70.0% | 2025-04-14 |
| 42 | Qwen3 235B-A22B | Alibaba | 68.9% | 2025-06-03 |
| 43 | Claude Sonnet 3.7 | Anthropic | 68.2% | 2025-02-24 |
| 44 | Qwen Max | Alibaba | 67.2% | 2025-04-01 |
| 45 | Qwen Plus | Alibaba | 65.3% | 2025-04-07 |
| 46 | Phi-4 | Microsoft | 64.9% | 2025-01-31 |
| 47 | DeepSeek-V3 | DeepSeek | 64.9% | 2025-01-27 |
| 48 | Grok-2 (Dec 2024) | xAI | 63.5% | 2025-02-17 |
| 49 | Qwen2.5-VL 72B Instruct | Alibaba | 63.2% | 2025-01-27 |
| 50 | Llama 4 Scout | Meta | 62.3% | 2025-04-08 |
| 51 | Gemini 1.5 Flash (Sep 2024) | 61.9% | 2025-01-27 | |
| 52 | Claude Sonnet 3.5 v2 | Anthropic | 57.0% | 2025-01-27 |
| 53 | Qwen Turbo | Alibaba | 56.2% | 2025-04-07 |
| 54 | Qwen2.5-32B | Alibaba | 56.1% | 2025-01-30 |
| 55 | GPT-4o (2024-08-06) | OpenAI | 53.3% | 2025-01-27 |
| 56 | GPT-4o mini | OpenAI | 52.6% | 2025-01-27 |
| 57 | Claude 3.5 Sonnet | Anthropic | 51.7% | 2025-01-27 |
| 58 | GPT-4o (2024-05-13) | OpenAI | 51.1% | 2025-01-27 |
| 59 | Mistral Large 2.1 | Mistral | 50.3% | 2025-02-25 |
| 60 | GPT-4o (2024-11-20) | OpenAI | 49.8% | 2025-02-05 |
| 61 | Llama 3.1-405B | Meta | 49.8% | 2025-01-27 |
| 62 | Mistral Small 3.1 | Mistral | 46.8% | 2025-03-18 |
| 63 | GPT-4 Turbo | OpenAI | 46.7% | 2025-02-27 |
| 64 | Claude Haiku 3.5 | Anthropic | 46.4% | 2025-03-12 |
| 65 | Mistral Large 2407 | Mistral | 44.8% | 2025-01-27 |
| 66 | Mistral Small 3 | Mistral | 44.8% | 2025-01-30 |
| 67 | Tulu 3 (Tülu 3) 70B | Ai2 | 42.7% | 2025-01-27 |
| 68 | Llama-3.3-70B-Instruct | Meta | 41.6% | 2025-01-27 |
| 69 | Gemini 1.5 Pro (May 2024) | 40.8% | 2025-01-27 | |
| 70 | Llama-3.2-90B-Vision-Instruct | Meta | 39.4% | 2025-01-27 |
| 71 | Qwen2-72B | Alibaba | 39.1% | 2025-01-27 |
| 72 | Claude 3 Opus | Anthropic | 37.5% | 2025-01-27 |
| 73 | Llama-3.1-70B-Instruct | Meta | 36.7% | 2025-01-27 |
| 74 | Gemma 2 27B | 27.9% | 2025-01-27 | |
| 75 | WizardLM 2 8x22B | Microsoft | 25.7% | 2025-01-27 |
| 76 | Yi-1.5-34B | 01.AI | 25.5% | 2025-01-27 |
| 77 | Gemini 1.5 Flash (May 2024) | 25.1% | 2025-01-27 | |
| 78 | Mistral Large | Mistral | 24.5% | 2025-01-27 |
| 79 | Mixtral 8x22B Instruct | Mistral | 24.2% | 2025-01-27 |
| 80 | GPT-4 | OpenAI | 23.0% | 2025-01-27 |
| 81 | Llama-3.1-8B-Instruct | Meta | 22.9% | 2025-01-27 |
| 82 | Hermes 2 Theta Llama-3 70B | Nous Research | 22.7% | 2025-01-27 |
| 83 | Llama 3-70B | Meta | 22.6% | 2025-01-27 |
| 84 | Gemma 2 9B | 21.0% | 2025-01-27 | |
| 85 | Claude 3 Sonnet | Anthropic | 18.2% | 2025-01-27 |
| 86 | phi-3-medium 14B | Microsoft | 17.6% | 2025-01-31 |
| 87 | GPT-3.5-turbo | OpenAI | 15.9% | 2025-01-27 |
| 88 | Ministral 8B Instruct | Mistral | 14.9% | 2025-01-27 |
| 89 | Claude Haiku 3 | Anthropic | 14.9% | 2025-01-27 |
| 90 | Ministral 3B | Mistral | 14.4% | 2025-01-27 |
| 91 | Claude 2 | Anthropic | 11.7% | 2025-01-27 |
| 92 | DBRX | Databricks | 11.7% | 2025-01-27 |
| 93 | Gemini 1.0 Pro | 11.2% | 2025-01-27 | |
| 94 | Mistral Nemo | Mistral | 10.8% | 2025-01-27 |
| 95 | Mixtral 8x7B Instruct v0.1 | Mistral | 10.0% | 2025-01-27 |
| 96 | DeepSeek LLM 67B | DeepSeek | 6.4% | 2025-01-27 |
| 97 | Llama 3-8B | Meta | 6.1% | 2025-01-27 |
| 98 | Yi-34B | 01.AI | 5.2% | 2025-01-27 |
| 99 | Mistral 7B Instruct v0.3 | Mistral | 3.7% | 2025-01-27 |
| 100 | Llama 2-70B | Meta | 3.3% | 2025-01-27 |