Benchmarks
Humanity's Last Exam
Models scored
51
Best score
0.46
Metric
Accuracy
Humanity's Last Exam is an AI evaluation tracked by Epoch AI, with 51 scored model versions on record. The scores here are reported by third parties rather than produced by Epoch AI running the evaluation itself.
Scored models span September 2024 to April 2026. The best score recorded is 0.46, measured as Accuracy. Models come from Google DeepMind, OpenAI, Meta AI, Anthropic, Moonshot, Google, Z.ai (Zhipu AI),Tsinghua University, Google DeepMind,Google and others.
Full record
- Score metric
- Accuracy
- Models scored
- 51
- Best score recorded
- 0.46
- Earliest model scored
- Sep 24, 2024
- Most recent model scored
- Apr 16, 2026
- Third-party reported
- Yes
Leaderboard
| 01 | gemini-3.1-pro-preview | Google DeepMind | 0.46 |
| 02 | gpt-5.4-pro-2026-03-05_unknown | OpenAI | 0.44 |
| 03 | muse-spark | Meta AI | 0.41 |
| 04 | gemini-3-pro-preview | Google DeepMind | 0.38 |
| 05 | gpt-5.4-2026-03-05_xhigh | OpenAI | 0.36 |
| 06 | claude-opus-4-7_unknown | Anthropic | 0.36 |
| 07 | claude-opus-4-6_max | Anthropic | 0.34 |
| 08 | gpt-5-pro-2025-10-06_unknown | OpenAI | 0.32 |
| 09 | gpt-5.2-2025-12-11_unknown | OpenAI | 0.28 |
| 10 | gpt-5-2025-08-07_high | OpenAI | 0.25 |
| 11 | gpt-5-2025-08-07_unknown | OpenAI | 0.25 |
| 12 | claude-opus-4-5-20251101_unknown | Anthropic | 0.25 |
| 13 | kimi-k2.5 | Moonshot | 0.24 |
| 14 | gpt-5.1-2025-11-13_unknown | OpenAI | 0.24 |
| 15 | gemini-2.5-pro-preview-06-05 | Google DeepMind | 0.22 |
| 16 | o3-2025-04-16_high | OpenAI | 0.2 |
| 17 | gpt-5-mini-2025-08-07_unknown | OpenAI | 0.19 |
| 18 | o3-2025-04-16_medium | OpenAI | 0.19 |
| 19 | claude-opus-4-6 | Anthropic | 0.19 |
| 20 | gemini-2.5-pro-exp-03-25 | Google DeepMind | 0.18 |
| 21 | o4-mini-2025-04-16_high | OpenAI | 0.18 |
| 22 | gemini-2.5-pro-preview-05-06 | Google DeepMind | 0.18 |
| 23 | o4-mini-2025-04-16_medium | OpenAI | 0.14 |
| 24 | claude-sonnet-4-5-20250929_unknown | Anthropic | 0.14 |
| 25 | gemini-2.5-flash-preview-04-17 | Google DeepMind | 0.12 |
| 26 | claude-opus-4-1-20250805_unknown | Anthropic | 0.12 |
| 27 | gemini-2.5-flash-preview-05-20 | Google DeepMind | 0.11 |
| 28 | claude-opus-4-20250514_unknown | Anthropic | 0.11 |
| 29 | gemini-3.1-flash-lite | 0.09 | |
| 30 | glm-4.5 | Z.ai (Zhipu AI),Tsinghua University | 0.08 |
| 31 | GLM-4.5-Air | Z.ai (Zhipu AI),Tsinghua University | 0.08 |
| 32 | o1-pro-2025-03-19 | OpenAI | 0.08 |
| 33 | claude-3-7-sonnet-20250219_unknown | Anthropic | 0.08 |
| 34 | o1-2024-12-17_unknown | OpenAI | 0.08 |
| 35 | claude-sonnet-4-20250514_unknown | Anthropic | 0.08 |
| 36 | gpt-5.1-2025-11-13_none | OpenAI | 0.07 |
| 37 | gemini-2.0-flash-thinking-exp-01-21 | Google DeepMind,Google | 0.07 |
| 38 | Llama-4-Maverick-17B-128E-Instruct | Meta AI | 0.06 |
| 39 | gpt-4.5-preview-2025-02-27 | OpenAI | 0.05 |
| 40 | gpt-4.1-2025-04-14 | OpenAI | 0.05 |
| 41 | gemini-1.5-pro-002 | Google DeepMind | 0.05 |
| 42 | mistral-medium-2505 | Mistral AI | 0.05 |
| 43 | amazon.nova-pro-v1:0 | Amazon | 0.04 |
| 44 | claude-3-5-sonnet-20241022 | Anthropic | 0.04 |
| 45 | amazon.nova-lite-v1:0 | Amazon | 0.04 |
| 46 | gpt-4o-2024-11-20 | OpenAI | 0.03 |
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