FrontierMath - Tiers 1-3: leaderboard
300 original math problems from undergraduate to early postdoc difficulty across major branches of modern mathematics, each requiring hours of expert effort to solve.
Metric: Accuracy (%, 290 problems). Source: epoch.ai. Source: openai.com. Source: openai.com. Saturation forecast: Around January 2027. 100 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.5 (xHigh) | 51.7 | #31 (GPT-5.5) |
| 2 | GPT-5.5 Pro (xHigh) | 51 | #17 (GPT-5.5 Pro) |
| 3 | GPT-5.4 Pro (xHigh) | 50 | #32 (GPT-5.4 Pro) |
| 4 | GPT-5.4 (xHigh) | 47.6 | #79 (GPT-5.4) |
| 5 | Claude Opus 4.8 (Max) | 47.24 | #37 (Claude Opus 4.8) |
| 6 | Claude Opus 4.7 (xHigh) | 43.79 | #49 (Claude Opus 4.7) |
| 7 | Claude Opus 4.6 (Max) | 40.7 | #63 (Claude Opus 4.6) |
| 8 | GPT-5.2 (xHigh) | 40.7 | #110 (GPT-5.2) |
| 9 | GPT-5.2 (High) | 40.34 | #110 (GPT-5.2) |
| 10 | Muse Spark | 39 | #64 |
| 11 | Kimi K2.6 | 38.97 | #104 |
| 12 | Gemini 3.5 Flash (High) | 38.97 | #60 (Gemini 3.5 Flash) |
| 13 | Claude Opus 4.6 (Non-reasoning) | 38.28 | #63 (Claude Opus 4.6) |
| 14 | Gemini 3 Pro (Preview) | 37.6 | #66 |
| 15 | Gemini 3.1 Pro (Preview) | 36.9 | #58 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=frontiermath-tiers-1-3 · How It Works · Data refreshed daily, snapshot 2026-10-20.