GPT 6.2 Astra shows surprisingly strong zero-shot crystal generation, a year after the task looked hopeless
xennygrimmato_ · x · 2026-11-19
Inspired by recent posts on frontier models generating protein sequences, AdeeshKolluru re-tested zero-shot generation of MSUN crystal structures and found GPT 6.2 Astra appears trained to be good at it — a task that looked mostly hopeless when he tried it over a year ago. He notes autoregressive models work well once trained on stable structures in formats like cifs/json, and that Fable generates many more duplicates, hurting uniqueness. An empirical AI-for-Science capability observation.
More from Models
- Gemini 4 Neon High matches xHigh scores at 42% lower cost — adonis_singh · 2026-11-19
- Google shipped Argon, Carbon, Hydrogen and Neon in 5 weeks of Gemini 4 iteration — Afinetheorem · 2026-11-19
- Gemini 4 Neon ships: #1 on AA Coding Agent Index, $3/$15 per 1M tokens — Kyrannio · 2026-11-19
- LLMs zero-shot beat fine-tuned world action models on 7 of 10 robot tasks — paigeinsf · 2026-11-19
- Cloudflare open-weights redesigned Clairic 27B and Clairic-flash 9B extraction models — michellechen · 2026-11-19
- Gemini Neon test: it names which "AI SEO courses" are rebranded legacy content and which to skip — gaganghotra_ · 2026-11-19