ResearchNovember 17, 2026
-11
❄️

Claude secretly learned protein language

Claude was trained extensively on protein sequences, predicts the next amino acid, and outperforms specialized protein LLMs.

#Anthropic#Claude#protein#research#benchmark
Claude kann plötzlich Proteine lesen
Share Article
🔥 What happened Anthropic's latest Claude models (Fable 5.2, Opus 5.5) predict the next amino acid in little-studied bacterial proteins with 54% accuracy – but drop to 9% on shuffled sequences. No other model tested (GPT, Gemini, seven open-weight models) shows this ability. 💡 Why it matters The 45-point gap between real and shuffled sequences proves Claude learned actual protein sequences from databases like UniProt, not just scientific text. Fable 5.2 even beats ProtGPT3-10B, the largest open protein language model, by a wide margin. ⚡ Our take Anthropic quietly slipped protein sequences into training data – a massive, unannounced advantage. Anyone still thinking frontier LLMs are just text machines hasn't been paying attention.
The title, summary and analysis of this item were produced automatically by an AI system and have not been editorially reviewed. They may contain errors, bias or omissions — when in doubt, read the linked original source.