AI 评测与排行榜，分别告诉你什么
分清评测基准、一次评测运行与排行榜，用它们回答各自适合的问题。
Model reports
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 report no-tools and with-tools figures as separate numbers.
评测信息与证据
- 来源
- Model reports
- 发布者
- 未收录
- 评测发布日期
- 2025-01-23
- 采集日期
- 未收录
- 模态
- 未收录
本站保留来源报告的结果，不代表独立复测。请先核对任务版本、评测协议、分数单位和报告日期；不同运行条件不能直接合并为统一排名。
原始资料
报告成绩 27
按来源中的观察顺序展示，顺序不代表排名。日期保留原始精度。
| 模型 | 数值 | 单位 | 协议 | 报告日期与口径 | 来源 |
|---|---|---|---|---|---|
| DeepSeek-V4-Pro | 7.7 | percent | non-think, pass@1, no tools | 2026-04-22 · document_publication | DeepSeek-V4-Pro |
| DeepSeek-V4-Pro | 34.5 | percent | think high, pass@1, no tools | 2026-04-22 · document_publication | DeepSeek-V4-Pro |
| DeepSeek-V4-Pro | 44.7 | percent | think high, pass@1, with tools | 2026-04-22 · document_publication | DeepSeek-V4-Pro |
| DeepSeek-V4-Pro | 37.7 | percent | think max, pass@1, no tools | 2026-04-22 · document_publication | DeepSeek-V4-Pro |
| DeepSeek-V4-Pro | 48.2 | percent | think max, pass@1, with tools | 2026-04-22 · document_publication | DeepSeek-V4-Pro |
| DeepSeek-V4-Pro-0813 | 60.0 | percent | max reasoning effort, temp 1.0, top_p 0.95, with tools | 2026-08-13 · document_publication | DeepSeek-V4-Pro-0813 |
| DeepSeek-V4-Pro-0813 | 42.7 | percent | max reasoning effort, temp 1.0, top_p 0.95, without tools | 2026-08-13 · document_publication | DeepSeek-V4-Pro-0813 |
| DeepSeek-V4-Flash | 34.8 | percent | think max, pass@1, no tools | 2026-06-25 · document_publication | DeepSeek-V4 |
| Gemini 2.5 Pro | 21.6 | percent | no tools | 2025-06-17 · document_publication | Gemini 2.5 Pro |
| Claude 4 Opus | 10.7 | percent | no tools, Scale leaderboard | 2025-06-17 · document_publication | Gemini 2.5 Pro |
| Claude 4 Sonnet | 7.8 | percent | no tools, Scale leaderboard | 2025-06-17 · document_publication | Gemini 2.5 Pro |
| o3 high | 20.3 | percent | no tools, Scale leaderboard | 2025-06-17 · document_publication | Gemini 2.5 Pro |
| Gemini 3.1 Pro | 44.4 | percent | Thinking (High), No tools | 2026-02-19 · document_publication | Gemini 3.1 Pro |
| Gemini 3.1 Pro | 51.4 | percent | Thinking (High), Search (blocklist) + Code | 2026-02-19 · document_publication | Gemini 3.1 Pro |
| Gemma 4 (31B) | 19.5 | percent | no tools | 2026-03-11 · document_publication | Gemma 4 (31B) |
| Gemma 4 (31B) | 26.5 | percent | with search | 2026-03-11 · document_publication | Gemma 4 (31B) |
| Kimi K3 | 56.0 | percent | reasoning=max, with general tools | 2026-06-13 · document_publication | Kimi K3 |
| Kimi K3 | 43.5 | percent | reasoning=max, without tools | 2026-06-13 · document_publication | Kimi K3 |
| Qwen3.5-397B-A17B | 28.7 | percent | thinking (temp 0.6, top_p 0.95, top_k 20), no tools | 2026-02-16 · document_publication | Qwen3.5-397B-A17B |
| Qwen3.5-397B-A17B | 48.3 | percent | thinking (temp 0.6, top_p 0.95, top_k 20), with tools | 2026-02-16 · document_publication | Qwen3.5-397B-A17B |
| Qwen3.5-397B-A17B | 37.6 | percent | thinking (temp 0.6, top_p 0.95, top_k 20), no tools, HLE-Verified dataset | 2026-02-16 · document_publication | Qwen3.5-397B-A17B |
| Hy4 preview | 55.4 | percent | w/ tools | 2026-08-28 · document_publication | Hy4 preview |
| Hy4 preview | 43.4 | percent | w/o tools | 2026-08-28 · document_publication | Hy4 preview |
| GLM-5.1 | 31.0 | percent | as reported | 2026-04-03 · document_publication | GLM-5.1 |
| GLM-5.2 | 40.5 | percent | text-only subset, temp 1.0, top_p 0.95, max gen 163840 tokens, GPT-5.5 (medium) judge | 2026-06-16 · document_publication | GLM-5.2 |
| GLM-5.3-Flash | 55.3 | percent | HLE w/ tools, full set, temp 1.0, top_p 0.95, max gen 163840 tokens, max context 300K with context management, GPT-5.6-luna (medium) judge | 2026-08-27 · document_publication | GLM-5.3-Flash |
| GLM-5 | 30.5 | percent | text-only subset, temp 1.0, top_p 0.95, GPT-5.2 (medium) judge | 2026-02-11 · document_publication | GLM-5 |