Task Performance on Sokoban
56.2Success RateClaude Opus 4.7
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Claude Opus 4.72026.05 | 56.2 | 5.04 | |
| Claude Sonnet 4.62026.05 | 51.6 | 5.65 | |
| Gemini 3.1 Pro Preview2026.05 | 39.1 | 5.63 | |
| GPT-5.2 Instant2026.05 | 35.2 | 9.02 | |
| Qwen3 235B2026.05 | 7 | 10.77 |