Versions

8 versions, plus 3 variants and configurations

Versions of OSWorld
VersionReleasedMetricModelsResults
OSWorld 2.0 v2026.08.08 Offline — Partial, max of 3 runsLatestSeptember 30, 2026OSWorld Partial Score11
OSWorld 2.0 v2026.08.08 Offline — PartialAugust 8, 2026OSWorld Partial Score631
OSWorld 2.0 v2026.08.08 — BinaryAugust 8, 2026OSWorld Strict Score11
OSWorld 2.0 v2026.08.08 — PartialAugust 8, 2026OSWorld Partial Score22
OSWorld 2.0 v2026.08.08 — StrictAugust 8, 2026OSWorld Strict Score11
OSWorld 2.0 v2026.06.24 — BinaryJune 26, 2026OSWorld Strict Score11
OSWorld 2.0 v2026.06.24 — PartialJune 26, 2026OSWorld Partial Score11
OSWorld Verified
3 variants of Verified
Variants of OSWorld Verified
VariantConfigurationReleasedMetricModelsResults
OSWorld Verified — 100 steps100-step limitJuly 28, 2025OSWorld-Verified success rate88
OSWorld Verified — 15 steps15-step limitJuly 28, 2025OSWorld-Verified success rate44
OSWorld Verified — 50 steps50-step limitJuly 28, 2025OSWorld-Verified success rate44
July 28, 2025OSWorld-Verified success rate1818

Recently reported results: OSWorld 2.0 v2026.08.08 Offline — Partial, max of 3 runs

Latest version: OSWorld 2.0 v2026.08.08 Offline — Partial, max of 3 runs

Recently reported results for OSWorld 2.0 v2026.08.08 Offline — Partial, max of 3 runs
ModelScoreSource
Gemini 4 Argon (highest thinking settings)69.2%deepmind.google for Gemini 4 Argon (highest thinking settings) on OSWorld 2.0 v2026.08.08 Offline — Partial, max of 3 runs (opens in a new tab)OSWorld Partial ScoreReported Evaluation methodologyGoogle DeepMindSelf-reportedRecord for Gemini 4 Argon on OSWorld 2.0 v2026.08.08 Offline — Partial, max of 3 runs