Back to Benchmarks
🧠

HLE

Humanity's Last Exam

2500 expert-vetted questions across mathematics, sciences, humanities

🧠 Knowledge & Reasoning
Best: 57.9%
10 Models Evaluated
Champion

Claude Opus 4.8

Anthropic

57.9%

Best Score

Model Rankings

All models ranked by score on HLE. Scores are relative to the best score of 57.9%.

#ModelProviderScorevs Best
1Claude Opus 4.8Anthropic
57.9%
Best
2Claude Opus 4.7
Elo 1500
Anthropic
54.7%
-3.2%
3Claude Opus 4.6
Elo 1503
Anthropic
53%
-4.9%
4GLM-5
Elo 1454
Zhipu AI
50.4%
-7.5%
5Claude Sonnet 4.6
Elo 1460
Anthropic
49%
-8.9%
6Gemini 3.1 Pro
Elo 1492
Google
45.8%
-12.1%
7Gemini 3 Pro
Elo 1486
Google
45.8%
-12.1%
8GPT-5.4 Pro
Elo 1480
OpenAI
43.1%
-14.8%
9DeepSeek V4 Pro (Max)DeepSeek
37.7%
-20.2%
10Qwen 3.5
Elo 1450
Alibaba
28.7%
-29.2%

Score Distribution

Claude Opus 4.8
57.9%
Claude Opus 4.7
54.7%
Claude Opus 4.6
53%
GLM-5
50.4%
Claude Sonnet 4.6
49%
Gemini 3.1 Pro
45.8%
Gemini 3 Pro
45.8%
GPT-5.4 Pro
43.1%
DeepSeek V4 Pro (Max)
37.7%
Qwen 3.5
28.7%

Run This Benchmark

Learn how to evaluate your own models on HLE.

HLE is a standardized benchmark for evaluating AI model capabilities.

To run this benchmark on your own models, visit the official benchmark page or leaderboard for detailed instructions, code, and evaluation scripts.

Data compiled from official model cards, public leaderboards, and independent evaluations.