Benchmarks · math
FrontierMath Tier 4 (v2)
Epoch AI's FrontierMath Tier 4, 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
68 models · 74 results · data as of 2026-10-29 · epoch.ai ↗
Leaderboard
GPT-6.1 Sol max
All results
74 rows| # | Model | Effort | FrontierMath Tier 4 (v2) | Harness |
|---|---|---|---|---|
| 1 | GPT-6.1 Sol | max | 100.0% | Epoch AI |
| 2 | GPT-6 Astra | high | 97.6% | Epoch AI |
| 3 | GPT-6 Astra | max | 97.6% | Epoch AI |
| 4 | GPT-6 Astra | xhigh | 97.6% | Epoch AI |
| 5 | GPT-6 Astra | medium | 97.6% | Epoch AI |
| 6 | Opus 5.5 | max | 95.0% | Epoch AI |
| 7 | GPT-6.1 Terra | max | 92.7% | Epoch AI |
| 8 | Fable 5 | max | 90.2% | Epoch AI |
| 9 | GPT-6 Sol | max | 90.0% | Epoch AI |
| 10 | Fable 5.1 | max | 87.8% | Epoch AI |
| 11 | GPT-6 Astra | low | 87.8% | Epoch AI |
| 12 | GPT-5.6 Sol | max | 82.9% | Epoch AI |
| 13 | GPT-6 Astra | none | 82.9% | Epoch AI |
| 14 | Sonnet 5.5 | max | 80.5% | Epoch AI |
| 15 | GPT-5.6 Sol Pro | max | 80.5% | Epoch AI |
| 16 | Gemini 4 Argon | high | 78.0% | Epoch AI |
| 17 | GPT-5.5 Pro | xhigh | 78.0% | Epoch AI |
| 18 | Opus 5 | max | 73.2% | Epoch AI |
| 19 | GPT-5.5 | xhigh | 72.5% | Epoch AI |
| 20 | GPT-5.6 Terra | max | 70.7% | Epoch AI |
| 21 | GPT-5.6 Luna | max | 61.0% | Epoch AI |
| 22 | GPT-5.4 Pro | xhigh | 58.5% | Epoch AI |
| 23 | Opus 4.8 | max | 56.1% | Epoch AI |
| 24 | GPT-6 Luna | max | 56.1% | Epoch AI |
| 25 | GPT-5.4 | xhigh | 49.0% | Epoch AI |
| 26 | Haiku 5.5 | max | 46.3% | Epoch AI |
| 27 | DeepSeek V4.1 Flash 1017 | max | 46.3% | Epoch AI |
| 28 | Muse Spark 1.3 | max | 46.3% | Epoch AI |
| 29 | GPT-5.2 Pro | xhigh | 46.0% | Epoch AI |
| 30 | Muse Spark 1.3 | xhigh | 41.5% | Epoch AI |