Claude Sonnet 4.6
#44 of 274 scored.
AnthropicFeb 17, 2026Closed
Capabilities Index
Range 149.9–154.5
152.2
Context
Takes text, image, file
1M
Input
Per 1M tokens
$3.00
Output
Per 1M tokens
$15
Benchmarks
How it scores.
- GPQA Diamond83.2%
54th of 186
- SWE-bench Verified75.2%
14th of 32
- ARC-AGI-260.4%
24th of 79
- SimpleQA Verified35.5%
49th of 77
- Mock AIME85.8%
69th of 176
- Terminal-Bench53.4%
12th of 35
Everything else Epoch has for it
- ARC-AGI86.5%
- DTBench83.1%
- OSWorld72.1%
- WeirdML66.1%
- FrontierMath-2025-02-28-Private56.8%
- DeepResearch Bench54.9%
- LMCA54.7%
- ProofBench45.0%
- APEX-Agents43.0%
- DeepSWE29.9%
- FrontierCode24.3%
- ExploitBench23.6%
- FrontierMath-Tier-4-2025-07-01-Private13.8%
- OSWorld 2.09.3%
- Chess Puzzles8.5%
- Mystery Game Puzzles7.5%
On the index
Around it.
Sources