Announcement
FineDNA dataset released for training AI models on DNA and text
FineDNA supplied Carbon-X with 302 billion mid-training tokens, plus supervised and reinforcement-learning data for DNA tasks.
TLDR
Carbon-X is a Qwen3-14B-based model that accepts raw DNA in a chat and replies in English. FineDNA holds 514.3 billion mid-training tokens, of which Carbon-X read 302 billion, plus 203,923 supervised fine-tuning conversations and 98,433 reinforcement-learning prompts. Hugging Face Bio reports that Carbon-X averaged 66.7% across five DNA benchmarks, ahead of the other open models it evaluated.
Combined views
137
13 Sources, first seen ago
