Benchmark
Humanity's Last Exam Benchmark
What is Humanity's Last Exam?
Tool access and test-time compute budget move this score more than model capability does, so two reported figures are rarely comparable. Cards increasingly…
- Released
- 2025-01-23
- Evaluates
- reasoning
- Openness
- unknown
- Importer
- Model reports
- Reported scores
- 32
Source provenance
- Original evidence https://lastexam.ai/
Humanity's Last Exam links
- Website https://lastexam.ai/
Humanity's Last Exam results and reported scores
Model reports
| Model | Organization | Reported value | Reported | Evidence |
|---|---|---|---|---|
| Claude 4 Opus | Anthropic | 10.7 | 2025-06-17 | evidence |
| Claude 4 Sonnet | Anthropic | 7.8 | 2025-06-17 | evidence |
| DeepSeek-V4-Flash | DeepSeek | 34.8 | 2026-06-25 | evidence |
| DeepSeek-V4-Pro | DeepSeek | 7.7 | 2026-04-22 | evidence |
| DeepSeek-V4-Pro | DeepSeek | 34.5 | 2026-04-22 | evidence |
| DeepSeek-V4-Pro | DeepSeek | 44.7 | 2026-04-22 | evidence |
| DeepSeek-V4-Pro | DeepSeek | 37.7 | 2026-04-22 | evidence |
| DeepSeek-V4-Pro | DeepSeek | 48.2 | 2026-04-22 | evidence |
| DeepSeek-V4-Pro-0813 | DeepSeek | 60.0 | 2026-08-13 | evidence |
| DeepSeek-V4-Pro-0813 | DeepSeek | 42.7 | 2026-08-13 | evidence |
| DeepSeek-V4.1-Flash | DeepSeek | 36.8 | 2026-09-10 | evidence |
| DeepSeek-V4.1-Flash | DeepSeek | 63.9 | 2026-09-10 | evidence |
| Gemini 2.5 Pro | 21.6 | 2025-06-17 | evidence | |
| Gemini 3.1 Pro | 44.4 | 2026-02-19 | evidence | |
| Gemini 3.1 Pro | 51.4 | 2026-02-19 | evidence | |
| Gemma 4 (31B) | 19.5 | 2026-03-11 | evidence | |
| Gemma 4 (31B) | 26.5 | 2026-03-11 | evidence | |
| GLM-5 | Z.ai | 30.5 | 2026-02-11 | evidence |
| GLM-5.1 | Z.ai | 31.0 | 2026-04-03 | evidence |
| GLM-5.2 | Z.ai | 40.5 | 2026-06-16 | evidence |
| GLM-5.3-Flash | Z.ai | 55.3 | 2026-08-27 | evidence |
| Hy4 preview | Tencent | 55.4 | 2026-08-28 | evidence |
| Hy4 preview | Tencent | 43.4 | 2026-08-28 | evidence |
| Kimi K3 | Moonshot AI | 56.0 | 2026-06-13 | evidence |
| Kimi K3 | Moonshot AI | 43.5 | 2026-06-13 | evidence |
| o3 high | OpenAI | 20.3 | 2025-06-17 | evidence |
| Qwen3.5-397B-A17B | Qwen | 28.7 | 2026-02-16 | evidence |
| Qwen3.5-397B-A17B | Qwen | 48.3 | 2026-02-16 | evidence |
| Qwen3.5-397B-A17B | Qwen | 37.6 | 2026-02-16 | evidence |
| Seed2.0 Lite | ByteDance | 27.8 | 2026-02-14 | evidence |
| Seed2.0 Mini | ByteDance | 21.5 | 2026-02-14 | evidence |
| Seed2.0 Pro | ByteDance | 32.4 | 2026-02-14 | evidence |
Scores are partitioned by the source that reported them and are never merged into a single cross-source ranking, because the sources measure different things and say so.
Which sources cite Humanity's Last Exam?
- DeepSeek-V4.1-Flash DeepSeek, 2026-09-10
- Hy4 preview Tencent, 2026-08-28
- GLM-5.3-Flash Z.ai, 2026-08-27
- DeepSeek-V4-Pro-0813 DeepSeek, 2026-08-13
- Qwen3.8-27B Qwen, 2026-08-05
- Grok 4.5 xAI, 2026-07-08
- GPT-5.6 OpenAI, 2026-06-26
- DeepSeek-V4 DeepSeek, 2026-06-25
- GLM-5.2 Z.ai, 2026-06-16
- Kimi K3 Moonshot AI, 2026-06-13
- Claude Fable 5 and Mythos 5 Anthropic, 2026-06-09
- DeepSeek-V4-Pro DeepSeek, 2026-04-22
22 source documents cite this benchmark.