evals.report
BenchmarksLabsCompareRun guidesIn the wild
BenchmarksAgents

OSWorld

OSWorld benchmarks multimodal AI agents on their ability to complete open-ended, real-world computer-use tasks (operating GUIs across web, files, and applications) in live operating-system environments via screenshots and mouse/keyboard control, measured by execution-based task success rate.

Agentstask success rateHigher is better
ScoresAboutRun this benchmark

What is OSWorld?

OSWorld benchmarks multimodal AI agents on their ability to complete open-ended, real-world computer-use tasks (operating GUIs across web, files, and applications) in live operating-system environments via screenshots and mouse/keyboard control, measured by execution-based task success rate. evals.report tracks reported OSWorld scores with the model, source, status, date, and run caveats attached — official leaderboard scores, vendor-reported launches, and clearly labeled community runs.

Top reported OSWorld score: Claude Fable 5 — 85.0% (task success rate).

ModelLabScore↓Source modelStatusDate
Claude Fable 5Anthropic85.0%Claude Fable 5VerifiedJun 9, 2026Details
Claude Opus 4.8Anthropic83.4%—VerifiedMay 28, 2026Details
Claude Opus 4.7Anthropic82.8%—VerifiedApr 16, 2026Details
Claude Sonnet 5Anthropic81.2%Claude Sonnet 5VerifiedJun 30, 2026Details
Muse Spark 1.1Meta80.8%Muse Spark 1.1VerifiedJul 9, 2026Details
Claude Mythos PreviewAnthropic79.6%—UnverifiedApr 7, 2026Details
GPT-5.5OpenAI78.7%—UnverifiedApr 23, 2026Details
Gemini 3.5 FlashGoogle DeepMind78.4%—UnverifiedMay 19, 2026Details
Gemini 3.1 Pro PreviewGoogle DeepMind76.2%—VerifiedFeb 19, 2026Details
GPT-5.4OpenAI75.0%—UnverifiedMar 5, 2026Details
Kimi K2.6OpenMoonshot AI73.1%—UnverifiedApr 20, 2026Details
Claude Opus 4.6Anthropic72.7%—UnverifiedFeb 5, 2026Details
Claude Sonnet 4.6Anthropic72.1%—UnverifiedFeb 17, 2026Details
MiniMax M3OpenMiniMax70.1%—UnverifiedJun 1, 2026Details
Claude Opus 4.5Anthropic66.3%—UnverifiedNov 24, 2025Details
GPT-5.3-CodexOpenAI64.7%—UnverifiedFeb 5, 2026Details
Claude Sonnet 4.5Anthropic61.4%—UnverifiedSep 29, 2025Details
GPT-5.2OpenAI47.3%—UnverifiedDec 11, 2025Details

Each row reports the model’s task success rate on OSWorld. Click a row for the full run context.