Agentic computer use

Agent's Last Exam

A hard set of desktop and operating-system tasks an AI agent has to finish by looking at the screen and working the machine itself. The score is the share it passes outright — partial credit does not count. Higher is better.

Rankings

Higher is better

Agent's Last Exam — frequently asked questions

What is Agent's Last Exam?
A hard set of desktop and operating-system tasks an AI agent has to finish by looking at the screen and working the machine itself. The score is the share it passes outright — partial credit does not count. Higher is better.
Which AI model scores highest on Agent's Last Exam?
Gemini 3.7 Flash by Google holds the best Agent's Last Exam result among tracked models, at 26.3% (released Aug 13 2026). Higher scores are better on this benchmark.
What are the top 1 models on Agent's Last Exam?
1. Gemini 3.7 Flash (Google) — 26.3%.
How many models have a published Agent's Last Exam score?
1 tracked model has a published Agent's Last Exam score. Scores are the figures reported by each lab at that model's release, so this page is a record of results over time rather than a re-run leaderboard.
← All benchmarks