About the Blender AI Arena

A community-driven benchmark for AI tools that create 3D content in and for Blender.

What Is This?

The Blender AI Arena is a benchmark site where users compare AI-generated 3D outputs in blind side-by-side matchups. You see two outputs, pick the better one (or tie/both bad), and your vote feeds into a live ranking system. No accounts needed.

Rating Methodology: Glicko-2

We use the Glicko-2 rating system, developed by Mark Glickman at Harvard. It extends the classic Elo system with two key additions:

Rating Deviation (RD) — measures how uncertain the rating is. New models start with high RD (350); as votes accumulate, RD drops toward 30.
Volatility — tracks whether a model's performance is consistent or erratic. Consistent performers earn more stable rankings.

The Conservative Rating (rating minus 2× RD) is used for ranking. This rewards models with many consistent votes over ones with high ratings from few matchups.

Initial Rating: 1500

Initial RD: 350

RD Floor: 30 | RD Ceiling: 350

Conservative Sigmas: 2

Volatility Tau: 0.5

Models in the Arena

The arena currently benchmarks 8 AI tools across three categories: 3D generators (Meshy, Tripo, Rodin), AI assistants (Claude, ChatGPT, Gemini, 3D-Agent), and legacy addons (BlenderGPT). See the full leaderboard.

Built by 3d-agent.com

The Blender AI Arena is built and maintained by 3d-agent.com — an AI agent platform that connects Claude, ChatGPT, and Gemini directly to Blender via MCP.

Want AI That Actually Creates in Blender?

3d-agent connects directly to Blender via MCP.

Try 3d-agent →