Hugging Face Releases Carbon-X AI Model for DNA Analysis

Hugging Face released Carbon-X on Thursday, an open-source AI model designed to allow scientists to chat with DNA using natural language. Developed in collaboration with the University of Naples Federico II and the Zhongguancun Academy, the model was built by training the Qwen3-14B language model on genomic sequences from humans, mammals, plants, fungi, and protozoa.

The model enables researchers to query DNA sequences to predict disease risks or generate new sequences for gene therapies. To prevent potential misuse, the developers explicitly excluded viruses from the training data. Carbon-X features 14 billion parameters, allowing it to run locally on user hardware to ensure sensitive patient data remains private and is not shared with external providers.

While the tool aims to democratize genomic research and advance personalized medicine, the developers noted that the model's predictions have not yet been validated through wet lab experiments.