Benchmarks
FrontierMath
Models scored
101
Best score
0.52
Metric
mean_score
FrontierMath is an AI evaluation tracked by Epoch AI, with 101 scored model versions on record. Epoch AI runs this evaluation directly, so the scores are reproducible against their published logs.
Scored models span June 2024 to May 2026. The best score recorded is 0.52, measured as mean_score. Models come from Anthropic, Alibaba, Google, Z.ai (Zhipu AI), Moonshot, OpenAI, Meta AI, Google DeepMind and others.
Full record
- Score metric
- mean_score
- Models scored
- 101
- Best score recorded
- 0.52
- Earliest model scored
- Jun 20, 2024
- Most recent model scored
- May 28, 2026
- Third-party reported
- No
Leaderboard
| 01 | gpt-5.5-pro-pre-release_high | OpenAI | 0.52 |
| 02 | gpt-5.5-pre-release_xhigh | OpenAI | 0.52 |
| 03 | gpt-5.5-pro-pre-release_xhigh | OpenAI | 0.51 |
| 04 | gpt-5.4-pro-2026-03-05_xhigh | OpenAI | 0.5 |
| 05 | gpt-5.4-2026-03-05_xhigh | OpenAI | 0.48 |
| 06 | claude-opus-4-8_max | Anthropic | 0.47 |
| 07 | claude-opus-4-7_xhigh | Anthropic | 0.44 |
| 08 | gpt-5.2-2025-12-11_xhigh | OpenAI | 0.41 |
| 09 | claude-opus-4-6_max | Anthropic | 0.41 |
| 10 | gpt-5.2-2025-12-11_high | OpenAI | 0.4 |
| 11 | claude-opus-4-6_32K | Anthropic | 0.4 |
| 12 | claude-opus-4-6_64K | Anthropic | 0.4 |
| 13 | muse-spark | Meta AI | 0.39 |
| 14 | gemini-3.5-flash_high | 0.39 | |
| 15 | kimi-k2.6 | Moonshot | 0.39 |
| 16 | claude-opus-4-6 | Anthropic | 0.38 |
| 17 | gemini-3-pro-preview | Google DeepMind | 0.38 |
| 18 | gemini-3.1-pro-preview | Google DeepMind | 0.37 |
| 19 | gpt-5.2-2025-12-11_medium | OpenAI | 0.37 |
| 20 | gemini-3-flash-preview | Google DeepMind | 0.36 |
| 21 | glm-5.1 | Z.ai (Zhipu AI) | 0.33 |
| 22 | gpt-5-2025-08-07_high | OpenAI | 0.32 |
| 23 | claude-sonnet-4-6_16K | Anthropic | 0.32 |
| 24 | gpt-5.1-2025-11-13_high | OpenAI | 0.31 |
| 25 | gemini-2.5-deep-think-2025-08-01-webapp | Google,Google DeepMind | 0.29 |
| 26 | gpt-5.4-mini-2026-03-17_high | OpenAI | 0.28 |
| 27 | fireworks/kimi-k2p5 | Moonshot | 0.28 |
| 28 | gpt-5-2025-08-07_medium | OpenAI | 0.27 |
| 29 | gpt-5-mini-2025-08-07_high | OpenAI | 0.27 |
| 30 | gpt-5.1-2025-11-13_medium | OpenAI | 0.27 |
| 31 | gpt-5.2-2025-12-11_low | OpenAI | 0.27 |
| 32 | qwen3.6-plus | Alibaba | 0.26 |
| 33 | gpt-5.4-nano-2026-03-17_high | OpenAI | 0.26 |
| 34 | o4-mini-2025-04-16_high | OpenAI | 0.25 |
| 35 | qwen3.6-max-preview | Alibaba | 0.23 |
| 36 | fireworks/deepseek-v3p2 | DeepSeek | 0.22 |
| 37 | moonshotai/Kimi-K2-Thinking | Moonshot | 0.21 |
| 38 | qwen3.5-plus | Alibaba | 0.21 |
| 39 | claude-opus-4-5-20251101 | Anthropic | 0.21 |
| 40 | claude-opus-4-5-20251101_32K | Anthropic | 0.21 |
| 41 | gpt-5-mini-2025-08-07_medium | OpenAI | 0.2 |
| 42 | claude-opus-4-5-20251101_16K | Anthropic | 0.2 |
| 43 | grok-4-0709 | xAI | 0.2 |
| 44 | o4-mini-2025-04-16_medium | OpenAI | 0.19 |
| 45 | o3-2025-04-16_high | OpenAI | 0.19 |
| 46 | gpt-5.1-2025-11-13_low | OpenAI | 0.17 |
| 47 | o3-2025-04-16_medium | OpenAI | 0.17 |
| 48 | glm-5 | Z.ai (Zhipu AI) | 0.16 |
| 49 | claude-sonnet-4-5-20250929_32K | Anthropic | 0.15 |
| 50 | gemini-2.5-pro | Google DeepMind | 0.14 |
| 51 | claude-sonnet-4-5-20250929_59K | Anthropic | 0.13 |
| 52 | o3-mini-2025-01-31_high | OpenAI | 0.12 |
| 53 | o4-mini-2025-04-16_low | OpenAI | 0.11 |
| 54 | gemini-2.5-pro-preview-06-05 | Google DeepMind | 0.1 |
| 55 | qwen3.6-flash | Alibaba | 0.1 |
| 56 | o3-2025-04-16_low | OpenAI | 0.1 |
| 57 | claude-sonnet-4-5-20250929 | Anthropic | 0.09 |
| 58 | o1-2024-12-17_high | OpenAI | 0.09 |
| 59 | Qwen/Qwen3-235B-A22B-Thinking-2507 | Alibaba | 0.08 |
| 60 | gpt-5-nano-2025-08-07_high | OpenAI | 0.08 |
More benchmarks
SourceEpoch AI, 'AI Benchmarking Hub'. Published online at epoch.ai. Retrieved 2026-07-29 from https://epoch.ai/benchmarks. Licensed under CC BY 4.0.
← All benchmarks