Claude Fable 5.2 beats protein models on little-studied proteins
First merged summary.
Fable 5.2 beat open protein models and other LLMs on 668 little-studied proteins, including 56 published in August 2026.
A researcher tested whether frontier models can continue little-studied bacterial and archaeal proteins, a task famous proteins cannot separate from memorized papers. Claude Fable 5.2 and Opus 5.5 predicted the next amino acid far above shuffled controls; 20 other language models did not. Fable 5.2 scored 54% on real sequences and 9% on shuffled ones, and beat ProGen3-3B and ProtGPT3-10B on the same left-to-right task. Opus 5.5 matched ProtGPT3-10B and trailed ProGen3-3B. An update says the models still predicted 56 proteins first published in August 2026.
- The test set has 668 proteins from 167 unknown-function Pfam families.
- Fable 5.2 scored 54% on real sequences and 9% on shuffled ones.
- Opus 5.5 matched ProtGPT3-10B and trailed ProGen3-3B.
- Accuracy rose with larger families, including families with no papers.
- GPT-6.2, Gemini and seven open-weight models stayed near zero.
What each outlet adds
beyond the shared facts- Hacker NewsScores, protein-model comparisons, and the August 2026 protein holdout.
- Organizations
- AnthropicContra LabsBenchFlow
Coverage
oldest first · one row per article · UTCBlog finds latest Claude models predict obscure protein sequences
Fable 5.2 beat open protein models on 668 little-studied proteins; GPT, Gemini and open-weight LLMs stayed near chance.
Two unnamed models lead Contra Labs design board ahead of Fable 5.2
The Frontier Design Benchmark used 33 client briefs, 68 designers and 4,311 blind votes.