OSWorldBrowse 296

OSWorld

Can the model plan, use tools, and finish a multi-step task?

Latest stable2.0 · Anthropic partial-score table
Agents3 ranked models3 reported values1 reportsHigher is better

Top rankings

One row per model, using its best reported score across effort settings.

3
RankModelBest scoreBest reported setting
1Claude Opus 5.5Anthropic81.8%max effortAnthropic report1 value · 1 reportSep 22, 2026 · source
2Claude Fable 5.1Anthropic80.7%Reported configurationAnthropic report1 value · 1 reportSep 22, 2026 · source
3Claude Opus 5Anthropic74.0%Reported configurationAnthropic report1 value · 1 reportSep 22, 2026 · source

Effort curve

Every sourced cost-linked effort value for this exact version. Lines connect complete sweeps only.

0
No cost-linked effort sweep for this version.
Definition and comparison boundarylimited methodology

Can the model plan, use tools, and finish a multi-step task?

Agentic execution across tools, environments, or long-running workflows. Higher is better. The value is the percentage reported in this lab's table.

Harness, tools, prompts, attempt count, and benchmark version can materially change the result. Compare within one reporting context.

Anthropic prints the exact partial-score values. Its table does not specify a task-release revision, trial count, or a more detailed agent scaffold for this comparison.

Method / source