Time horizon plot update with judgment-heavy tasks
METR updated its time horizon plot with longer and more open-ended tasks and found exponential growth continues, while models do worse on tasks that need judgment.
The earlier plot scored models by how long human testers took on tasks the models could complete. The update adds tasks that take humans far longer and tasks that are more open-ended. MIT Technology Review reports models are weaker on judgment-heavy tasks than on coding. The Information reports a doubling time of 3.6 months and that recent models, including OpenAI's GPT-6 Astra, can complete tasks that take a human about a work week.
- Date
- Friday 20 November 2026
- Lab
- METR
- Kind
- paper
- Access
- paper only
Figures
| Measure | Value | Measured by |
|---|---|---|
| capability doubling time | 3.6 months as reported by The Information; MIT Technology Review describes the original plot as doubling roughly every four months | METR |
This entry draws on press coverage of METR's report, mainly MIT Technology Review and The Information.
Sources
- technologyreview.com/2026/11/20/1149399/the-most-misunderstood-graph-in-ai-just-got-a-majo
- theinformation.com/briefings/ais-can-now-outperform-humans-week-length-tasks-ai-nonprofit-
- techmeme.com/261120/p25#a261120p25
This record was partly confirmed: some claims could not be checked on 20 November 2026. How we check