Multimodal models are now using Blender to generate 3D assets

giffmana · x · 2026-07-25

Multimodal models using Blender are now producing convincing 3D assets

The post says it was unthinkable two years ago that multimodal language models would use Blender as the best path to generate 3D assets — yet that is now happening.

The quoted example says Claude Opus 5 outperformed Fable on a Blender MCP castle-building task, which surprised the author enough to call the gap “ridiculous.”

Original post →

More from Multimodal

Multimodal channel →