Anthropic
Claude Mythos 5
6 benchmarks
GPQA Diamond
#7 of 118
94.1%
BrowseComp
#7 of 45
88.0%
Terminal-Bench 2.1
#5 of 27
88.0%
SWE-bench Pro
#2 of 43
80.3%
ExploitBench
#1 of 2
78.0%
Toolathlon
#1 of 7
61.7%
benchmark.space canonical public ledger of AI capability