Search COMPUTER USE

Search the evidence-backed index by name, vendor, or use case.

COMPUTER USE / BENCHMARK

OSWorld Verified: methodology & results

Understand OSWorld Verified Verified revision · 2025-07-28: Task success under versioned execution checks, evaluation environment and comparability limits.

Release
Verified revision · 2025-07-28
Environment
desktop
Owner
XLang / OSWorld research team
Metric
Task success under versioned execution checks
Original methodology ↗

What this evaluation measures

Owner announces the Verified upgrade on July 28, 2025. Historical OSWorld scores must remain version-separated.

Definition refers to the July 2025 Verified revision, not OSWorld 1.0 or V2. Exact task denominator and run configuration must be supplied for every future result.

OSWorld Verified evaluates desktop tasks. Its reported metric is task success under versioned execution checks. Read the release-specific task definitions before treating this as a proxy for your own workflows.

Results belong to the exact system and evaluation setup named in the source. A vendor’s current API is not automatically the system used in a historical run. Hosted browser reachability, task completion and full desktop interaction are different measurements.

Owner methodology and measurement boundary

Owner announces the Verified upgrade on July 28, 2025. Historical OSWorld scores must remain version-separated.

Reproducibility requirements

Resolve exact task denominator, release and run configuration for each result.

Version history and interpretation

Verified was announced July 28, 2025. OSWorld V2 introduces a separate long-horizon task family; neither name permits migration of scores from earlier evaluations.

Attributed results

Numeric results are withheld: a version-matched result has not passed our evidence checks. This methodology guide remains useful for understanding scope and reproducibility.

Can these results be compared?

No unified ranking is published. These records do not establish a controlled comparison across current tools.

  • Match benchmark code, tasks, assets and environment versions.
  • Check whether resets, retries, task exclusions and human interventions are counted.
  • Compare the same metric: task success is different from reachability or cost per attempt.
  • Re-run representative tasks using your own constraints before a purchase or deployment decision.
RESEARCH NOTES

Sources & verification

Reviewed Sep 28, 2026. Architecture recommendations are editorial analysis; linked vendor documentation supports the underlying capability and safety facts.

  1. github.com/xlang-ai/OSWorld
How we verify evidence →

Key facts & source context

Owner announces the Verified upgrade on July 28, 2025. Historical OSWorld scores must remain version-separated.

Scope: desktop; Definition refers to the July 2025 Verified revision, not OSWorld 1.0 or V2. Exact task denominator and run configuration must be supplied for every future result. · Unit: Task success under versioned execution checks · Version: Verified revision · 2025-07-28

supported · benchmark owner · historical · Verified 2026-09-28 · XLang / OSWorld research team

Limitation: Definition refers to the July 2025 Verified revision, not OSWorld 1.0 or V2. Exact task denominator and run configuration must be supplied for every future result.