LuminaBench: leaderboard

A model in its vendor's coding agent builds one voxel pagoda, one 3D seaplane scene and one Mario Kart-style game, and the number is the mean of the three build ratings. Each build is rated on LuminaBench's own rating scale by the method recorded for it: rubric reviews with pairwise matchups, a rubric gameplay assessment, or the site owner's relative placement, none of them blinded. Unlike WebDev Arena's crowd votes over many prompts, each configuration is one build per test.

Metric: Lumina Capabilities Index (mean owner rating of three builds). Source: luminabench.com. Saturation forecast: Around December 2027. 10 models tracked. Updated 16 Nov 2026.

Top models

#ModelSettingScoreOverall rank
1Claude Fable 5.2High1116#1
2Claude Opus 5.5Max1099#5
3Claude Sonnet 5.5Max1096#19
4GPT-6 AstraMax1085#6
5Gemini 4 ArgonHigh1078#12
6GPT-6.1 SolMax1075#10
7GPT-6.1 TerraMax1064#21
8Grok 4.7xHigh1009#47
9Claude Haiku 5.5High1007#70
10Gemini 3.8 FlashLow854#36

No result here: #2 Gemini 4 Carbon.

Interactive version: theaggregate.ai/benchmark?slug=luminabench · How It Works · Data refreshed daily, snapshot 2026-11-16.