Atlas

Benchmarks

← All benchmarks

BlenderWorld Multimodal

Code · 2026-10-14

The 145 BlenderWorld 3D creation tasks that include image or video references to reproduce. A coding agent works in an isolated GPU container with shell, file and image tools plus a Blender MCP server, and must deliver an editable .blend project with the requested renders or video. A GPT-5.6 Terra judge scores each task as the equal-weight mean of a weighted binary-rubric pass rate and eight 1-10 visual-dimension ratings, both on 0-100; a missing .blend scores zero. The score is the mean fused task score over these tasks.

Top models (higher is better)

ModelScore
GPT-6 Astra80.7
Claude Opus 572.1
DeepSeek V4.1 Flash65.5
GPT-5.6 Sol65.5
GLM-5.3 Flash57.8
Opus 4.856.1
Grok 4.655.4
Kimi K349.4
Qwen3.8 27B48.6
Gemini 3.8 Flash33.2
Loading Atlas data…