Claude Sonnet 4.6 — Real Scores Across 7 Benchmarks (2026)

Anthropic

Claude Sonnet 4.6

7 benchmarks

AIME

#20 of 39

94.0%

GPQA Diamond

#11 of 49

89.9%

SWE-bench Verified

#8 of 40

79.6%

MMLU Pro

#21 of 29

79.2%

BrowseComp

#11 of 17

74.0%

OSWorld

#6 of 16

72.5%

Terminal-Bench 2.0

#9 of 14

59.1%