Blender AI Leaderboard
Live rankings by community votes. Models rated using the Glicko-2 system — the same algorithm behind chess and esports rankings.
| # | Model | Provider | Type | Rating | Conservative | Win Rate | W-L-D | Confidence |
|---|---|---|---|---|---|---|---|---|
| 1 | 3D-Agent | 3D-Agent | agent | 1660 | 1560 | 53.3% | 240-90-120 | 94% |
| 2 | Claude | Anthropic | assistant | 1640 | 1530 | 51.2% | 220-100-110 | 92% |
| 3 | Meshy AI | Meshy | 3d-generator | 1620 | 1490 | 48.8% | 195-120-85 | 89% |
| 4 | ChatGPT | OpenAI | assistant | 1610 | 1490 | 48.2% | 200-115-100 | 91% |
| 5 | Tripo AI | Tripo | 3d-generator | 1580 | 1440 | 45.3% | 170-130-75 | 88% |
| 6 | Gemini | assistant | 1570 | 1420 | 44.3% | 155-125-70 | 86% | |
| 7 | Rodin AI | Hyper3D | 3d-generator | 1550 | 1390 | 43.8% | 140-115-65 | 84% |
| 8 | BlenderGPT | Community | addon | 1450 | 1270 | 34.6% | 90-130-40 | 81% |
3,480 total matchups · Updated in real-time
Rating Methodology
The Blender AI Arena uses the Glicko-2 rating system, which extends the classic Elo system with rating deviation (uncertainty) and volatility tracking.
- Initial Rating: 1500 (all models start equal)
- Conservative Rating: Rating minus 2× deviation — rewards consistency
- Confidence: Decreases deviation from 350 to 30 as more votes accumulate
Models are ranked by conservative rating to penalize uncertainty. New models with few votes will rank lower until the community establishes a consensus.
Want AI That Actually Creates in Blender?
3d-agent connects directly to Blender via MCP — generating, modifying, and rendering 3D models with natural language.
Try 3d-agent →