Autonomy of AI assistants on extended tasks
How long a task an AI agent can complete independently. Today about 17 hours of professional work.
↑ A rising value means progress. The dots measure the distance travelled from the starting point in 2020 towards the 2040 mark, not the raw reading. Logarithmic scale: each doubling counts equally. For values above 16 hours, METR states that its measurements remain uncertain.
Not a forecast: 65 points do not mean the 2040 mark will be reached at a particular time. The points measure the distance covered since the 2020 start, not the measurement itself. And a single indicator says little; only together do the 37 form a picture.
Save this indicator. The home page then shows how it has moved since you saved it.
Stored only in your browser. No account, no data with us.
The research institute METR measures how long a task may take if an AI agent completes it independently in half of all attempts, benchmarked against the time a professional needs. In 2020 that figure was about 9 seconds; today the best model manages about 17 hours. Since 2023 the value has doubled roughly every four months. The scale is therefore logarithmic: each doubling counts equally. 100 points corresponds to a working year (2,000 hours).
"From about one working month (167 hours, roughly 80 points), AI completes entire projects on its own. That is the turning point from tool to colleague."
From the baseline 0.14 (2020) the indicator is moving towards 120'000. Current status: 1'044.78. This corresponds to 65 of 100 Points.
Source: METR, Time Horizons — 50% time horizon of the best measured model · As at 9 October 2026 · metr.org/time-horizons