Claude Mythos Preview — Real Scores Across 13 Benchmarks (2026)

Anthropic

Claude Mythos Preview

13 benchmarks

USAMO 2026

#1 of 3

97.6%

GPQA Diamond

#1 of 49

94.6%

SWE-bench Verified

#1 of 40

93.9%

SWE-bench Multilingual

#1 of 3

87.3%

BrowseComp

#2 of 17

86.9%

CyberGym

#1 of 3

83.1%

Terminal-Bench 2.0

#1 of 14

82.0%

GraphWalks BFS

#1 of 1

80.0%

OSWorld

#1 of 16

79.6%

SWE-bench Pro

#1 of 13

77.8%

Humanity's Last Exam

#1 of 24

64.7%

SWE-bench Multimodal

#1 of 2

59.0%

Humanity's Last Exam

#1 of 24

56.8%