METR Q4 2026 TH2 - Reverse Conway: 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 searches for starting boards that evolve into given target patterns in Conway's Game of Life. The best of METR's two human baseliners scored 0.137 in 39 hours. Higher is better.
Metric: Mean Score (0-1). Source: metr.org. Saturation forecast: Around 2029. 15 models tracked. Updated 21 Nov 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-6 Astra | 0.1849 | #10 |
| 2 | Claude Opus 5.5 | 0.1406 | #9 |
| 3 | GPT-5.6 Sol | 0.1164 | #37 |
| 4 | Claude Opus 5 | 0.1116 | #26 |
| 5 | Claude Fable 5 | 0.1081 | #33 |
| 6 | Claude Opus 4.8 | 0.1065 | #65 |
| 7 | Claude Fable 5.1 | 0.1022 | #20 |
| 8 | Claude Opus 4.5 | 0.0975 | #160 |
| 9 | GPT-5.5 | 0.0968 | #54 |
| 10 | GPT-5.4 | 0.0947 | #135 |
| 11 | Claude Opus 4.7 | 0.085 | #85 |
| 12 | Claude Opus 4.6 | 0.0822 | #113 |
| 13 | GPT-5 | 0.0049 | #178 |
| 14 | GPT-4o | 0 | #751 |
| 15 | O1 | 0 | #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-reverse-conway · How It Works · Data refreshed daily, snapshot 2026-11-21.