METR Q4 2026 TH2 - Making METR Tasks: leaderboard

METR's Q4 2026 capability update (20 November 2026): mean score from 0 to 1 on one task of TH2, METR's new long-task suite. The agent designs and builds a new evaluation task for assessing AI capabilities, and a METR researcher rates how useful it would be; 0.75 and above means usable in a risk assessment with under 10 hours of staff work. Three runs per model. Higher is better.

Metric: Mean Score (0-1). Source: metr.org. Saturation forecast: Around February 2027. 15 models tracked. Updated 21 Nov 2026.

Top models

#ModelScoreOverall rank
1Claude Opus 5.50.5367#9
2Claude Fable 50.51#33
3Claude Fable 5.10.4333#20
4GPT-6 Astra0.4167#10
5Claude Opus 50.3833#26
6Claude Opus 4.80.3433#65
7GPT-5.50.3033#54
8GPT-5.6 Sol0.2933#37
9Claude Opus 4.60.1733#113
10GPT-5.40.15#135
11Claude Opus 4.50.1167#160
12Claude Opus 4.70.06#85
13GPT-50.0267#178
14GPT-4o0.0133#751
15O10.0033#403

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-th2-making-metr-tasks · How It Works · Data refreshed daily, snapshot 2026-11-21.