Benchmarks / Reasoning & Math
MATH Level 5
Hardest tier of the MATH competition dataset.
maintainer unknown · 108 observations on record
Versions
1 version
| Version | Tasks |
|---|---|
| MATH Level 5 current math-level-5 | — |
Current leaderboard
MATH Level 5 · Accuracy · top 25 of 108 systems
| # | System | Accuracy % |
|---|---|---|
| 1 | GPT 5 (2025-08-07)high effort OpenAI | 98.1% |
| 2 | GPT 5 (2025-08-07)medium effort OpenAI | 97.9% |
| 3 | GPT 5 Mini (2025-08-07)high effort OpenAI | 97.8% |
| 4 | O4 Mini (2025-04-16)high effort OpenAI | 97.8% |
| 5 | O3 (2025-04-16)high effort OpenAI | 97.8% |
| 6 | Claude Sonnet 4.5 2025092932K budget Anthropic | 97.7% |
| 7 | Qwen3 Max (2025-09-23) Alibaba | 97.1% |
| 8 | GPT 5 Mini (2025-08-07)medium effort OpenAI | 96.8% |
| 9 | Deepseek R1 0528 DeepSeek | 96.6% |
| 10 | O3 Mini (2025-01-31)high effort OpenAI | 96.5% |
| 11 | Claude Haiku 4.5 2025100132K budget Anthropic | 96.4% |
| 12 | Gemini 2.5 Pro Preview 05.06 Google DeepMind | 95.9% |
| 13 | Gemini 2.5 Pro Preview 03.25 Google DeepMind | 95.6% |
| 14 | GPT 5 Nano (2025-08-07)medium effort OpenAI | 95.2% |
| 15 | O3 Mini (2025-01-31)medium effort OpenAI | 95.2% |
| 16 | GPT 5 Nano (2025-08-07)high effort OpenAI | 94.9% |
| 17 | O1 (2024-12-17)high effort OpenAI | 94.7% |
| 18 | O1 (2024-12-17)medium effort OpenAI | 94.4% |
| 19 | Deepseek R1 DeepSeek | 93.0% |
| 20 | Claude 3.7 Sonnet 2025021964K budget Anthropic | 91.2% |
| 21 | Grok 3 Mini Betalow effort xAI | 90.9% |
| 22 | Claude 3.7 Sonnet 2025021932K budget Anthropic | 90.0% |
| 23 | Deepseek R1 Distill Llama 70B DeepSeek | 89.9% |
| 24 | O1 Mini (2024-09-12)high effort OpenAI | 89.2% |
| 25 | Grok 3 Beta xAI | 88.8% |
One row per evaluated system — reasoning-effort variants rank separately.