Benchmarks · math
FrontierMath Tiers 1–3 (v2)
Epoch AI's FrontierMath Tiers 1 3, version 2: expert-written mathematics problems evaluated with Python tools. Accuracy on the private set, not the public sample questions. Data: Epoch AI, CC BY.
Category
86 models · 119 results · data as of 2026-10-29 · epoch.ai ↗
Leaderboard
GPT-6 Astra max
All results
119 rows| # | Model | Effort | Score | Harness |
|---|---|---|---|---|
| 1 | GPT-6 Astra | max | 93.7% | Epoch AI |
| 2 | GPT-6.1 Sol | max | 93.7% | Epoch AI |
| 3 | Opus 5.5 | max | 91.2% | Epoch AI |
| 4 | GPT-6.1 Terra | max | 90.9% | Epoch AI |
| 5 | Fable 5.1 | max | 90.2% | Epoch AI |
| 6 | GPT-6 Sol | max | 89.8% | Epoch AI |
| 7 | GPT-5.6 Sol | max | 89.1% | Epoch AI |
| 8 | Sonnet 5.5 | max | 88.8% | Epoch AI |
| 9 | GPT-5.5 Pro | xhigh | 87.7% | Epoch AI |
| 10 | Fable 5 | max | 87.0% | Epoch AI |
| 11 | Gemini 4 Argon | high | 86.0% | Epoch AI |
| 12 | GPT-5.6 Terra | max | 86.0% | Epoch AI |
| 13 | Opus 5 | max | 85.6% | Epoch AI |
| 14 | GPT-5.5 | xhigh | 85.3% | Epoch AI |
| 15 | GPT-5.4 Pro | xhigh | 82.5% | Epoch AI |
| 16 | GPT-5.6 Luna | max | 82.1% | Epoch AI |
| 17 | Opus 4.8 | max | 80.0% | Epoch AI |
| 18 | GPT-6 Luna | max | 78.9% | Epoch AI |
| 19 | GPT-5.4 | xhigh | 78.6% | Epoch AI |
| 20 | Haiku 5.5 | max | 75.1% | Epoch AI |
| 21 | Muse Spark 1.3 | xhigh | 74.4% | Epoch AI |
| 22 | Muse Spark 1.3 | max | 74.0% | Epoch AI |
| 23 | GPT-5.2 Pro | xhigh | 74.0% | Epoch AI |
| 24 | Kimi K3 | max | 72.2% | Epoch AI |
| 25 | Gemini 3.7 Flash | high | 71.6% | Epoch AI |
| 26 | DeepSeek V4.1 Flash 1017 | max | 71.2% | Epoch AI |
| 27 | Opus 4.7 | max | 70.2% | Epoch AI |
| 28 | GLM 5.3 | max | 68.8% | Epoch AI |
| 29 | Gemini 3.8 Flash | high | 68.4% | Epoch AI |
| 30 | GPT-5.2 | xhigh | 67.4% | Epoch AI |