VoxelCodeBench - Visual Quality: leaderboard

VoxelCodeBench (2026) evaluates 3D spatial reasoning through code generation: from minimal API documentation and a few examples, a model writes Python that builds voxel scenes in an Unreal Engine renderer, and human annotators judge the rendered output. This board is the mean human rating of the output's visual quality on a 0 to 10 scale; higher is better.

Metric: Visual quality (Look): mean human rating of the rendered output's aesthetic quality on a 0-10 scale; higher is better. Source: arxiv.org. Saturation forecast: Around 2030. 8 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1GPT-55.71#178
2Claude Sonnet 4.55.01#261
3GPT-5 Mini4.86#364
4Claude Opus 44.13#326
5GPT-5 Chat3.66#449
6Claude 3 Opus3.4#1431
7Claude 3.5 Sonnet3.3#752
8Gemini 3 Pro1.36#145

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=voxelcodebench-visual-quality · How It Works · Data refreshed daily, snapshot 2026-11-21.