Anthropic's latest Claude models, specifically Fable 5.2 and Opus 5.5, demonstrate a surprising ability to predict protein sequences, a task typically handled by specialized protein language models. This capability was tested by providing the models with the beginning of a protein sequence and evaluating their accuracy in predicting the next amino acid, both on intact and shuffled sequences. The Claude models significantly outperformed other frontier LLMs, including GPT and Gemini, on this task, suggesting they were trained on protein sequence data at scale. AI
IMPACT Demonstrates potential for LLMs to be trained on specialized scientific data, opening avenues for AI in biological research.
RANK_REASON The item details a novel capability discovered in existing LLMs through a specific benchmark, rather than a new model release or a significant industry event. [lever_c_demoted from research: ic=1 ai=1.0]
Read on HN — claude cli stories →
- Claude
- Claude Opus 4.7
- Claude Opus 5.5
- DeepSeek V4.1 Flash
- ESM-2
- Gemini
- GPT
- GLM 5.3
- GPT-6 Astra
- GPT-OSS 120B
- Kimi K3.1 Pro
- Minimax M3
- Nemotron 3 Ultra
- PG-LLM
- ProGen3
- ProtGPT3
- ProtGPT3-10B
- Qwen 3.8
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →