Leaderboard PLANTS
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. 340 votes cast.
Image→3D reconstruction
One or more photos reconstructed into a 3D mesh.
- 1 Pixal3D 1427
- 1 TRELLIS 2 1370
- 1 Meshy 6 1370
12 models ranked · 0 of 12 firm
19 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 openai/gpt-5.6-sol-pro 2112
- 1 x-ai/grok-4.5 1695
- 1 openai/gpt-5.6-sol 1412
17 models ranked · 0 of 17 firm
20 more votes firms the next model here
View board →Text→3D (native)
A text prompt turned directly into 3D.
- 1 Meshy v6 text 1295
- 1 Tripo H3.1 text 1152
- 1 Tripo P1 text 1136
7 models ranked · 0 of 7 firm
14 more votes firms the next model here
View board →Agentic 3D
An agent renders, critiques, and revises its own 3D model in a loop.
- 1 openai/gpt-5.6-sol (agentic) 1470
- 1 google/gemini-3.1-pro-preview (agentic) 1262
- 1 anthropic/claude-opus-4.8 (agentic) 1113
13 models ranked · 0 of 13 firm
21 more votes firms the next model here
View board →Filters & bias audit
Bias audit — left(A) win rate 0.569 (≈0.50 = unbiased) · tie 0.059 · bad 0.191
VLM judge (Sonnet 4.6, multi-view) — automated LLM-judge rankings by paradigm
Loading automated rankings…