VoxelCodeBench - Material Correctness: leaderboard
Metric: Material correctness (%): share of tasks whose rendered output uses the specified materials and colors, judged by a human annotator; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 8 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5 | 87.9 |
| 2 | Claude Sonnet 4.5 | 84.6 |
| 3 | GPT-5 Mini | 82.4 |
| 4 | GPT-5 Chat | 75.9 |
| 5 | Claude Opus 4 | 72.1 |
| 6 | Claude 3.5 Sonnet | 70.3 |
| 7 | Claude 3 Opus | 54.8 |
| 8 | Gemini 3 Pro | 18.8 |
Interactive version: theaggregate.ai/benchmark?slug=voxelcodebench-material-correctness · How It Works · Data refreshed daily, snapshot 2026-10-07.