AnalysisAI ModelsNovember 2, 2026

FineDNA dataset teaches any LLM DNA, powers Carbon-X

Read original source →x.com

FineDNA packs 300B+ tokens of genomic sequences and DNA-rich text for midtraining, plus ~200K SFT conversations and ~98K verifiable RL prompts. Training Qwen3-14B on it produced Carbon-X, described as the strongest open model on DNA; code and recipe are open.

2 sources

More stories today

Open the live feed