Astra beats Sol but not SOTA on hard wet-lab biology, citing scarce public data
nlarusstone · x · 2026-09-05
The author notes Astra outperforms Sol but is still not SOTA on hard wet-lab biology tasks. The bottleneck: most biological reasoning isn't captured in public data, making it hard to train for. They remain excited about the progress.
More from Models
- GPT-6 builds a 9-stage interactive explainer on looped transformers — cto_junior · 2026-09-06
- Blogger pegs 30% odds OpenAI already solved Navier-Stokes, 50% partial progress — scaling01 · 2026-09-05
- LLMs write locally coherent but globally incoherent quests: an MMO writer's thousands-of-quests problem — HLCYSWAP · 2026-09-05
- GPT-6 'Astra' at capacity? User burns 9% of weekly quota in 20 hours then hits rate limit — sick_burns2000 · 2026-09-05
- GPT Astra High Shows Best-Ever 3D Understanding in 40-Minute Sculpting Test — rms80 · 2026-09-05
- Why Agents Last Exam Scores Jumped: Computer Use, Bigger Models, Diverse RL Envs — dejavucoder · 2026-09-05