DuelLab Overall: leaderboard
DuelLab evaluates model-generated game-playing programs by compiling submitted code and running head-to-head tournaments on hidden abstract strategy games.
Metric: DuelLab Score. Source: benchmarks.duellab.org. Status: years away from saturation. 121 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.6 Sol (xHigh) | 87.3 |
| 2 | Claude Opus 5 (xHigh) | 76 |
| 3 | Claude Opus 5 (Medium) | 73.5 |
| 4 | GPT-5.4 (xHigh) | 72.7 |
| 5 | Claude Opus 5 | 68.5 |
| 6 | GPT-5.5 (xHigh) | 67.5 |
| 7 | Claude Opus 4.8 (xHigh) | 66.6 |
| 8 | GPT-5.6 Sol (Medium) | 66.5 |
| 9 | Claude Fable 5 (Medium) | 66.1 |
| 10 | Gemini 3.6 Flash (High) | 63.2 |
| 11 | Grok 4.5 (High) | 62.1 |
| 12 | GPT-5.6 Terra (xHigh) | 57.5 |
| 13 | GPT-5.6 Sol | 55.8 |
| 14 | Gemini 3.6 Flash (Medium) | 55.8 |
| 15 | Gemini 3.5 Flash (Medium) | 55.4 |
Interactive version: theaggregate.ai/benchmark?slug=duellab-overall · How It Works · Data refreshed daily, snapshot 2026-09-05.