Leaderboard ALL KINGDOMS
Each generation method is ranked on its own board — scores aren't comparable across methods (they come from separate match pools). Pick a method to compare the models within it. 1621 votes cast.
Image→3D reconstruction
One or more photos reconstructed into a 3D mesh.
- 1 Meshy 6 1146
- 1 TRELLIS 2 1067
- 2 Hunyuan3D v3 1044
12 models ranked · 10 of 12 firm
2 more votes firms the next model here
View board →LLM procedural (code-gen)
An LLM writes code (e.g. Blender) that builds the 3D model.
- 1 z-ai/glm-4.6v 1013
- 1 openai/gpt-5.6-sol 1013
- 1 x-ai/grok-4.5 1013
5 models ranked · 0 of 5 firm
28 more votes firms the next model here
View board →Text→3D (native)
A text prompt turned directly into 3D.
- 1 Meshy v6 text 1086
- 1 Tripo H3.1 text 1064
- 1 Tripo P1 text 1038
7 models ranked · all firm
View board →Agentic 3D
An agent renders, critiques, and revises its own 3D model in a loop.
- 1 z-ai/glm-4.6v (agentic) 1013
- 1 openai/gpt-5.6-sol (agentic) 1013
- 1 x-ai/grok-4.5 (agentic) 1013
4 models ranked · 0 of 4 firm
29 more votes firms the next model here
View board →Filters & bias audit
Bias audit — left(A) win rate 0.499 (≈0.50 = unbiased) · tie 0.07 · bad 0.163
VLM judge (Sonnet 4.6, multi-view) — automated LLM-judge rankings by paradigm
Loading automated rankings…