Recent evaluations presented frontier models with the starting segments of 668 obscure proteins across 167 Pfam families lacking solved structures or close UniProt relatives. Claude Fable 5.2 and Claude Opus 5.5 successfully completed the sequences, while competing models from GPT, Gemini, DeepSeek, and Qwen showed near zero accuracy. Fable 5.2 achieved 54% accuracy on real sequences compared to 9% on shuffled baselines, even surpassing dedicated open protein models. Meanwhile, genomic testing indicated that Opus 5.5 can predict human and mouse coding sequences but struggles with the rest of the genome.
Key points
The ability to predict undocumented protein families shows these models learned broad bio…: The ability to predict undocumented protein families shows these models learned broad biological patterns rather than simply memorizing academic texts.
Performance in biological sequence completion scales with family size even in the total a…: Performance in biological sequence completion scales with family size even in the total absence of published research papers.
Success remains unevenly distributed across biology because models struggle with non-codi…: Success remains unevenly distributed across biology because models struggle with non-coding regions and bacterial genes despite mastering human coding sequences.
Sources checked by our analysis pipeline: 1
A next-amino-acid test on little-studied proteins finds that the latest Claude models were trained on protein sequences at scale. Other frontier models show no sign of it.
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.
Frontier LLMs have started to look good at protein tasks. Atharv Jayprakash
Famous proteins can’t tell the two apart: GFP is printed in countless papers, and most frontier models recite it. So we tested proteins that almost nobody writes about.
We give a model the start of a protein and ask for the next amino acids. Then we ask again on…
International research highlights how conversational tools are becoming integrated into the emotional and social routines of modern youth. Young demographics increasingly treat these platforms as relational partners rather than mere utility instruments. This shift marks a profound transition in global socialization patterns among adolescents and young adults.
Lexicographers have identified a cultural pushback against automated technology, marked by an emerging linguistic trend prioritizing authentic human-made work. This commercial and creative pivot gained urgency after sophisticated language models began reliably deceiving users in conversational benchmarks. Consequently, creators, lawmakers, and brands are taking active measures to protect and verify original human labor.
Search engines are increasingly placing synthesized summaries above original links, absorbing independent publisher material directly into their interfaces. This shift drastically diminishes referral traffic for specialized informational resources that rely on search visibility to sustain their operations. Consequently, digital creators who provide civic and local guidance are facing severe revenue losses that threaten their long-term financial viability.