# AI agents turn photos into editable 3D Blender code

> Researchers evaluated six GPT models on a new benchmark and found the best model hit about 53% accuracy on indoor scenes, but geometry judgments remain near chance.

Oossa · 2026-10-03 · https://oossa.com/en/ai-agents-turn-photos-into-editable-3d-blender-code

A team introduced LEGO‑Bench, a test set of 208 photos rendered from simulated 3D scenes, to measure how well coding agents can write Blender scripts from a single image. Six GPT configurations were run; the top model, GPT‑6 Astra, reconstructed indoor scenes correctly 53.4% of the time and outdoor scenes 39.6%. The agents could usually produce a usable scene file, but their ability to judge geometric accuracy was close to random. Adding a plug‑in that anchors the scene to the reference image boosted weaker models by up to 62.7%.

## The facts

- GPT‑6 Astra scored 53.4% on indoor scenes and 39.6% on outdoor scenes in LEGO‑Bench.
- The new LEGO‑Plugin improved the weakest models by up to 62.7%.

## Why it matters

For developers of 3D tools, the results show that AI can automate initial scene creation, but reliable geometry checks are still needed.

## Sources & references

1. [AI agents build 3D scenes from photos but have no idea if they got it right](https://the-decoder.com/ai-agents-build-3d-scenes-from-photos-but-have-no-idea-if-they-got-it-right/) – The Decoder, 2026-10-03

Last updated: 2026-10-03