FrontierMath - Tiers 1-3 — leaderboard
Metric: Accuracy (%, 290 problems). Source: epoch.ai. 100 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.5 (xHigh) | 51.7 |
| 2 | GPT-5.5 Pro (xHigh) | 51 |
| 3 | GPT-5.4 Pro (xHigh) | 50 |
| 4 | GPT-5.4 (xHigh) | 47.6 |
| 5 | Claude Opus 4.8 (Max) | 47.24 |
| 6 | Claude Opus 4.7 (xHigh) | 43.79 |
| 7 | GPT-5.2 (xHigh) | 40.7 |
| 8 | Claude Opus 4.6 (Max) | 40.7 |
| 9 | GPT-5.2 (High) | 40.34 |
| 10 | Claude Opus 4.6 (Thinking 32K) | 40 |
| 11 | Muse Spark | 39 |
| 12 | Kimi K2.6 | 38.97 |
| 13 | Gemini 3.5 Flash (High) | 38.97 |
| 14 | Claude Opus 4.6 (Non-reasoning) | 38.28 |
| 15 | Gemini 3 Pro (Preview) | 37.6 |
Interactive version: theaggregate.ai/benchmark?slug=frontiermath-tiers-1-3 · How the rankings work · Data refreshed daily, snapshot 2026-07-22.