Benchmarks › Humanity's Last Exam
Humanity's Last Exam
Answer expert questions using text only
61.4top score · Claude Opus 5.5
334models ranked
224open-weights models
Pass@1 accuracyhigher is better
Oct 20, 2026last updated
How the test works
1
Task
2,158 questions
- What the model doesAnswer expert questions using text only
- Kinds of tasksExpert knowledge questions
- DomainsMathematics, humanities, natural sciences
2
Run
1 run per task
- Token budgetNon-reasoning: up to 16,384 output tokens, adjusted to model limits; reasoning: model-specific maximum
- Temperature0 for non-reasoning; 0.6 for reasoning, unless the model lab recommends otherwise
3
Score
Pass@1 (solved on the first try) accuracy
- Graded byGPT-5.6 Luna (medium) equality-checker judge
- Better isHigher
Sources: artificialanalysis.ai
Humanity's Last Exam leaderboardtop 10 of 334
Full leaderboard with filters →| Model | Humanity's Last Exam | Per task | Price | Speed | Supports | Context |
|---|---|---|---|---|---|---|
61.4max | $5.98 | $4 / $20 | 96 | 1M | ||
59.1max | $7.63 | $10 / $50 | 69 | 1M | ||
57.1high | $1.99 | $2 / $10 | 84 | 1M | ||
55.5max | $8.75 | $10 / $50 | 67 | 1M | ||
55max | $5.46 | $2 / $10 | 137 | 1M | ||
54.9max | $5.86 | $5 / $25 | 55 | 1M | ||
54.7max | $3.26 | $10 / $50 | 69 | 1.05M | ||
53.6 | — | — | 1M | |||
52.9max | $0.72 | $2 / $10 | 52 | 1.05M | ||
50.1max | $4.52 | $4 / $20 | 161 | 1M |
Humanity's Last Exam questions
What is Humanity's Last Exam?›
Answer expert questions using text only. It has 2,158 questions. Created by Center for AI Safety and Scale AI.
Which model is best on Humanity's Last Exam?›
Claude Opus 5.5 leads Humanity's Last Exam with 61.4, ahead of Claude Fable 5.1 (59.1) and Gemini 4 Argon (57.1).
What is the best open-weights model on Humanity's Last Exam?›
MiMo v2.6 Pro with 49.4, then Kimi K3 (46.9).
How is Humanity's Last Exam scored?›
Pass@1 (solved on the first try) accuracy; gpt-5.6 luna (medium) equality-checker judge; 1 run per task per task.