Claude Sonnet 4
#108 of 274 scored.
AnthropicMay 22, 2025Closed
Capabilities Index
Range 138.9–143.0
141.7
Context
Takes image, text, file
200K
Input
Per 1M tokens
$3.00
Output
Per 1M tokens
$15
Benchmarks
How it scores.
- GPQA Diamond72.3%
83rd of 186
- ARC-AGI-25.9%
49th of 79
- Humanity's Last Exam3.1%
31st of 40
- Mock AIME71.1%
89th of 176
Everything else Epoch has for it
- MATH level 584.4%
- Lech Mazur Writing81.4%
- METR Time Horizons62.0%
- DTBench61.8%
- Aider polyglot61.3%
- Fiction.LiveBench46.9%
- DeepResearch Bench46.6%
- WeirdML46.1%
- OSWorld43.9%
- ARC-AGI40.0%
- GeoBench37.0%
- Cybench35.0%
- SimpleBench34.6%
- LMCA34.1%
- The Agent Company33.1%
- FrontierMath-2025-02-28-Private7.3%
- GSO-Bench4.9%
- VPCT1.0%
- FrontierMath-Tier-4-2025-07-01-Private0.0%
On the index
Around it.
Sources