Create

Sign in to ReadmeX

or

How we handle your data: Privacy Policy

ReadmeX

Your AI reading companion

What would you like to catch up on?

Pick a briefing, or ask your own question.

Your community and people briefings stay personal to you.

Story

Claude models show sequence-aware protein prediction

AI summary

Researchers reported that Claude Fable 5.2 and Claude Opus 5.5 predicted amino-acid order in little-studied bacterial and archaeal proteins better than the other tested frontier models. Fable 5.2 scored 54% on the next amino acid for real sequences versus 9% for shuffled sequences and outperformed every open protein language model evaluated. Opus 5.5 scored 34.9% on real sequences versus 7.5% on shuffled sequences and was level with ProtGPT3-10B when the sequence-order gap was compared. Anthropic’s official introduction does not describe its training data, so the claim that these models were trained on protein sequences at scale remains an inference rather than a confirmed fact.

In-depth insight

The findings suggest that Fable 5.2 may contain sequence-specific biological knowledge and outperform specialized open protein language models. They also raise questions about the composition of frontier-model training data, although Anthropic has not confirmed the proposed explanation.

The latest Claude models were trained on protein sequences, at scale. Give Claude Fable 5.2 or Claude Opus 5.5 the start of a little-studied bacterial protein and they predict the next amino acid. Shuffle the protein and they can’t. No other model we tested can do this. Like a protein language model, they do better on bigger protein families, and they do it even in families with no papers at all. Fable 5.2 beats every open protein language model we ran, and Opus 5.5 is level with ProtGPT3-10B, the largest of them.

Update, 18 November 2026. Since the first version:

  • Proteins published after the training cutoff. With Leo Tianlai Chen, 56 proteins first made public in August 2026, with no earlier relative at 50% identity or more. The latest Claude models still predict them (Appendix G).
  • More protein language models. ProGen3-3B is the strongest open one we ran, and Fable 5.2 still beats it; Prot-xLSTM is in Appendix C.
  • More GPT models. GPT-6.2 Sol and Astra show no sign of it either.

Motivation

Frontier LLMs have started to look good at protein tasks. Atharv Jayprakash showed GPT-6 Astra ranking SOD1 variants as well as ESM-2, and PG-LLM found the same across many proteins. Did these models learn from protein sequence databases such as UniProt, or from what is written about proteins?

#Anthropic#Claude Fable 5.2#Claude Opus 5.5#UniProt#ProGen3-3B#ProtGPT3-10B#ESM-2#GPT-6.2 Sol#Gemini 4 Carbon

Original · 15 min readRead the original
Group chat

No comments yet. Start the conversation.