Frontier Log

Benchmark · InactiveSuperseded by OSWorld 2.1

OSWorld 2.0 (v2026.08.08)

Created by XLANG Lab, 8 Aug 2026 · Higher is better · Website

0102030405060Jul 2026Oct 202641.7%Fable 5.139.6%Opus 536.1%Fable 50102030405060Jul 2026Oct 202641.7%Fable 5.139.6%Opus 536.1%Fable 5

Settings and run setups

Task setScoringProviders
Offline setPartial scoreOpenAIGoogle
Anthropic's corrected setPartial scoreAnthropic
Anthropic's corrected setStrict pass rateAnthropic
Full set (108 tasks)Not statedMiniMax

Download image

Results

ModelScoreSource
Claude Fable 5.11 Sep 202641.7%Introducing Claude Fable 5.1 and Claude Mythos 5.1Corroborated The same number is printed in 1 other lab document: 1tasks where safeguards intervened scored zeroOur capture
Claude Opus 524 Jul 202639.6%Introducing Claude Fable 5.1 and Claude Mythos 5.1Corroborated The same number is printed in 1 other lab document: 1re-run on the August 2026 task release · reported laterOur capture
Claude Fable 59 Jun 202636.1%Introducing Claude Fable 5.1 and Claude Mythos 5.1Corroborated The same number is printed in 1 other lab document: 1re-run on the August 2026 task release · tasks where safeguards intervened scored zero · reported laterOur capture

All benchmarks