Astra results on my hobby Blender Benchmark

A developer tested GPT-6-Astra on a personal Blender benchmark project, inspired by MineBench. The results, primarily from one-shot "write a Python script" runs, showed a striking difference compared to other models. While multi-turn runs with Blender MCP were too expensive for large-scale testing, they yielded even more impressive results. The developer invites others to view and vote on the benchmark results.

This report is among the first to offer a direct comparison of GPT-6-Astra's performance against other LLMs on a Blender benchmark.