FrontierMath - Tiers 1-3 (v2) — leaderboard
Metric: Accuracy (%, 285 private v2 problems). Source: epoch.ai. 38 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.6 Sol (Max) | 89.12 |
| 2 | GPT-5.5 Pro (xHigh) | 87.72 |
| 3 | Claude Fable 5 (Max) | 87.02 |
| 4 | GPT-5.6 Terra (Max) | 85.96 |
| 5 | GPT-5.5 (xHigh) | 85.26 |
| 6 | GPT-5.4 Pro (xHigh) | 82.46 |
| 7 | GPT-5.6 Luna (Max) | 82.11 |
| 8 | Claude Opus 4.8 (Max) | 80 |
| 9 | GPT-5.4 (xHigh) | 78.6 |
| 10 | Kimi K3 (Max) | 72.18 |
| 11 | Claude Opus 4.7 (Max) | 70.18 |
| 12 | GPT-5.2 (xHigh) | 67.4 |
| 13 | Claude Opus 4.6 (Max) | 65.96 |
| 14 | Claude Sonnet 5 (Max) | 65.61 |
| 15 | Qwen 3.7 Max | 64.56 |
Interactive version: theaggregate.ai/benchmark?slug=frontiermath-tiers-1-3-v2 · How the rankings work · Data refreshed daily, snapshot 2026-07-22.