Visual Puzzle Solving on Visual Puzzle Tasks 1.0 (test)
100Hexagon Color Pattern Match AccuracyGPT-5 high
Evaluation Results
| Method | Links | ||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| GPT-5 highMultiple-choice options=provided for tasks with *2025.11 | 100 | 100 | 95.8 | 98 | 98.5 | 35.4 | 98 | 66.7 | 100 | 82 | 60 | 98 | 85 | 86.8 | |
| Gemini 2.5 ProMultiple-choice options=provided for tasks with *2025.11 | 98 | 94 | 87.5 | 100 | 94.9 | 83.3 | 84 | 83.7 | 56 | 54 | 58 | 100 | 67 | 81.5 | |
| Sora-2Multiple-choice options=false2025.11 | 96 | 94 | 85.4 | 52 | 81.9 | 45.8 | 58 | 51.9 | 56 | 66 | 44 | 64 | 57.5 | 66.2 | |
| Claude Sonnet 4.5Multiple-choice options=provided for tasks with *2025.11 | 92 | 100 | 62.5 | 66 | 80.1 | 93.8 | 46 | 69.9 | 86 | 88 | 54 | 100 | 82 | 78.8 |