MATH Level 5: leaderboard
The hardest tier (Level 5) of the Hendrycks MATH dataset - 1,324 competition-level problems spanning algebra, number theory, geometry, and combinatorics. Evaluated by Epoch AI.
Metric: Accuracy (%). Source: epoch.ai. Saturation forecast: Estimated already saturated. 108 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5 (2025-08-07) (High) | 98.13 | |
| 2 | GPT-5 (Medium) | 97.92 | #98 (GPT-5) |
| 3 | GPT-5 Mini (2025-08-07) (High) | 97.85 | #172 (GPT-5 Mini (2025-08-07)) |
| 4 | O4 Mini (2025-04-16) (High) | 97.83 | #179 (O4 Mini (2025-04-16)) |
| 5 | O3 (2025-04-16) (High) | 97.77 | #121 (O3 (2025-04-16)) |
| 6 | Claude Sonnet 4.5 (Thinking 32K) | 97.73 | #142 (Claude Sonnet 4.5) |
| 7 | Qwen 3 Max (2025-09-23) | 97.13 | #255 |
| 8 | GPT-5 Mini (2025-08-07) (Medium) | 96.79 | #172 (GPT-5 Mini (2025-08-07)) |
| 9 | DeepSeek R1 0528 | 96.64 | #222 |
| 10 | O3 Mini (2025-01-31) (High) | 96.49 | #240 (O3 Mini (2025-01-31)) |
| 11 | Gemini 2.5 Pro (Preview 05-06) | 95.9 | #147 |
| 12 | Gemini 2.5 Pro (Preview 03-25) | 95.56 | #123 |
| 13 | GPT-5 Nano (2025-08-07) (Medium) | 95.24 | |
| 14 | O3 Mini (Medium) | 95.17 | #272 (O3 Mini) |
| 15 | GPT-5 Nano (2025-08-07) (High) | 94.9 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=math-level-5 · How It Works · Data refreshed daily, snapshot 2026-10-20.