The standard for computer-use agents

How well can AI use a computer?

Computer Use is the industry reference for agents that operate real software. We track the scores that matter, explain how each was measured, and keep every benchmark in one place.

108OSWorld 2.0 tasks
~1.6hmedian human time per task
79.38%top partial score
—benchmarks indexed
OSWorld 2.0 leaderboard

Best agents to use computers

Long-horizon, real-world desktop workflows. Partial score is the standard: it measures how far an agent gets through work that takes a person over an hour and a half.

Ranked by partial score

From the official OSWorld 2.0 leaderboard, full task set. Partial score is our headline metric because it credits real progress on long tasks that few agents finish outright.

Official leaderboard
#ModelSetup Cost / task Reproduce

Rows mix two releases (v2.1 and v2026.08.08). Compare scores within the same release.

Reproduce links and trajectories are the ones each team submitted to the official board. Source: osworld-v2.xlang.ai, updated November 14, 2026.

What it measures
Whether an agent can finish complex professional workflows across real desktop apps from start to finish, and how far it gets when it fails.
Metrics
Partial (headline): share of each task's checkpoints reached. Binary: the task is fully complete.
Origin