METR Q4 2026 Time Horizon - TH1.1 + TH2 (50%): leaderboard

METR's Q4 2026 capability update (20 November 2026): the length of task, in hours of expert human work, that a model completes 50% of the time, fitted across tasks of different lengths. This is the post's headline series: the 222 older Time Horizon 1.1 tasks pooled with TH2, METR's new suite of 36 longer tasks (up to 200 human hours) scored against several human (score, time) points per task, for the 15 models run on both. Unbounded; higher is better. Reasoning effort and scaffold are not stated per model.

Metric: 50% Time Horizon (hours). Source: metr.org. Saturation forecast: Around December 2027. 15 models tracked. Updated 21 Nov 2026.

Top models

#ModelScoreOverall rank
1GPT-6 Astra40.27#10
2Claude Opus 5.538.96#9
3Claude Fable 5.130.21#20
4Claude Opus 519.91#26
5Claude Fable 515.58#33
6GPT-5.6 Sol11.94#37
7Claude Opus 4.811.84#65
8GPT-5.511.07#54
9Claude Opus 4.78.38#85
10Claude Opus 4.66.4#113
11GPT-5.45.94#135
12Claude Opus 4.54.59#160
13GPT-53.01#178
14O10.5875#403
15GPT-4o0.1185#751

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=metr-q4-2026-time-horizon-th1-1-plus-th2-50 · How It Works · Data refreshed daily, snapshot 2026-11-21.