Hugging Face Bio releases FineDNA, the open DNA-and-text training dataset behind Carbon-X (Apache-2.0)

FineDNA is the training data of Carbon-X, the model that taught Qwen3-14B to read DNA, for all three training stages: mid-training (514.30B tokens across 23 source configs, of which the run read 302B), SFT (203,923 conversations, 230.3M tokens) and RL (98,433 prompts, 61.1M tokens). Apache 2.0 and ungated on Hugging Face; announced on X on 1 November 2026 (repository created 30 October).