LuminaBench - Voxel: leaderboard
The rating of one voxel pagoda built by the model in its coding agent; most builds were rated by rubric reviews and pairwise matchups, the rest by the site owner's relative placement, none blinded. One build per configuration.
Metric: Owner rating of one voxel-pagoda-v1 build. Source: luminabench.com. Saturation forecast: Around January 2028. 10 models tracked. Updated 16 Nov 2026.
Top models
| # | Model | Setting | Score | Overall rank |
|---|---|---|---|---|
| 1 | Claude Fable 5.2 | High | 1114 | #1 |
| 2 | Claude Opus 5.5 | Max | 1095 | #5 |
| 3 | GPT-6 Astra | Max | 1092 | #6 |
| 4 | Claude Sonnet 5.5 | Max | 1090 | #19 |
| 5 | Gemini 4 Argon | High | 1087 | #12 |
| 6 | GPT-6.1 Sol | Max | 1081 | #10 |
| 7 | GPT-6.1 Terra | Max | 1077 | #21 |
| 8 | Claude Haiku 5.5 | High | 1044 | #70 |
| 9 | Grok 4.7 | xHigh | 980 | #47 |
| 10 | Gemini 3.8 Flash | High | 866 | #36 |
No result here: #2 Gemini 4 Carbon.
Interactive version: theaggregate.ai/benchmark?slug=luminabench-voxel · How It Works · Data refreshed daily, snapshot 2026-11-16.