METR Q4 2026 Time Horizon - Lower Lifecycle Cost Tasks (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. Fitted only on the half of the pooled TH1.1 + TH2 tasks that are cheaper than their length predicts to build, run and score, by a model's cost estimate: the tasks METR expects are easier to turn into training environments. 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 Astra69.64#10
2Claude Opus 5.568.62#9
3Claude Fable 5.158.86#20
4Claude Opus 531.4#26
5Claude Fable 518.63#33
6Claude Opus 4.816.1#65
7GPT-5.6 Sol14.09#37
8GPT-5.513.05#54
9Claude Opus 4.712.21#85
10Claude Opus 4.611.85#113
11GPT-5.411.02#135
12Claude Opus 4.57.41#160
13GPT-55.83#178
14O11.32#403
15GPT-4o0.1891#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-lower-lifecycle-cost-tasks-50 · How It Works · Data refreshed daily, snapshot 2026-11-21.