GPT-6 Astra crushes Claude in Blender tasks — the gap is training data, not intelligence
bookwormengr · x · 2026-09-08
A developer building AI tools inside Blender found that given the same brief (a 12-second forest path animation), GPT-6 Astra's output clearly beats Claude Fable 5.1. His explanation: Anthropic simply hasn't put much Blender data into its training corpus — the model isn't dumb, Blender just isn't part of its distribution yet.
Another engineer extended the point: Astra's Blender/CAD strength comes from training data exposure, not architecture or scale — echoing Karpathy's observation that early models were terrible at chess until chess data was added. He expects other models to catch up within 2-3 months.
Related event: Flagship Models Diverge Sharply in Same-Prompt Blender Test(2 posts)→
More from Models
- OpenAI ships GPT-6.1 Terra; Brockman says its computer use now beats his own speed — gdb · 2026-10-27
- Kimi K3.1 Pro regresses on Debate Benchmark while DeepSeek 1017 gains at lower cost — teortaxesTex · 2026-10-27
- 4 Hours of LLM Startup Research Produced a Convincing but Nonsense Product Thesis — annetgriffin · 2026-10-27
- Dev argues Claude burns far more tokens than GPT models for the same coding work — haider1 · 2026-10-27
- MiniMax matches Claude Opus 4.5 in non-benchmaxxed evals, users report — teortaxesTex · 2026-10-27
- Real-world math: GPT-6.1 Terra's half price is offset by 2-3x token usage, costing more than Sol — lordpuddingcup · 2026-10-27