VoxelBench Image Rating
Games
VoxelBench image rating is the Glicko-2 leaderboard track for image-capable models creating voxel builds. Ratings come from head-to-head human votes over rendered voxel constructions.
Top models (higher is better)
| Model | Score |
|---|---|
| Gemini 3.1 Pro Preview | 1715 |
| Opus 4.6 | 1472 |
| Gemini 3 Pro Preview | 1465 |
| GPT-5.4 | 1445 |
| GPT-5.2 | 1410 |
| Opus 4.7 | 1310 |
| Opus 4.5 | 1300 |
| GPT-5 | 1179 |
| GPT-5-Codex | 1170 |
| Gemini 2.5 Pro | 1168 |
| GPT-5.1 | 1109 |
| Opus 4.1 | 1098 |
| o3 | 969 |
| Haiku 4.5 | 870 |