Google reports 81.5% for Gemini 4 Carbon on OSWorld 2.1, a partial score on an 82-task offline subset, taking the best of three runs.
Tasks: 82-task offline (no-internet) subset; partial score.
Setting: Google says its results use the Gemini API at the highest thinking setting unless noted; no numeric budget is stated.
Limits: Provider-reported: Google publishes the result for its own model; it has not been independently reproduced here. Google says the scores are as of October 2026 for the model it calls Gemini 4 Carbon; equivalence to the build now served on every surface is not established. Results under the same benchmark name can be shown together as reported scores, but unmatched settings and missing methods rule out a controlled ranking. Best of three runs is not a single-run score.
Google DeepMind · Google DeepMind · Google DeepMind
