Frontier Log

Benchmark · No longer reported

τ²-bench

020406080100Jan 2026Apr 202699.3%Opus 4.6020406080100Jan 2026Apr 202699.3%Opus 4.6

Results

ModelScoreSourceHuman-verified
Claude Opus 4.65 Feb 2026 · Anthropic99.3%Claude Opus 4.6Source imageYes

Benchmark

github.comHigher is better

SierraIntroduced 9 Jun 2025

All benchmarks