Controlling Blender with Gemini 3

Gemini 3 was recently announced. According to the benchmarks, it belongs to the top tier in coding capabilities. I tried using it with the Blender API, and it felt not just competitive, but potentially better than other models. Here is a quick introduction.

Showcase

It generates objects like these from simple prompts:

It can even generate rigs and animations.

Since it seemed quite powerful, I built a Streamlit app to interactively input prompts.

https://github.com/unclepomedev/BlenderGeminiAgent

Conclusion

There are no specific benchmarks for Blender coding, so this is just my personal impression, but Gemini 3 feels superior to other models for this task. If this intuition is correct, I wonder if it is due to the influence of YouTube training data or its multimodal capabilities.