Hugging Face’s Carbon-X Brings AI to DNA Analysis and Gene Therapy Research

Hugging Face is releasing Carbon-X, an AI model the company says can analyze and generate DNA sequences. The model combines genomic data with plain language, allowing researchers to ask questions about DNA in a more conversational way. Lewis Tunstall, a senior research scientist at Hugging Face, described the tool as a way for scientists to essentially “chat with DNA.”

One example cited in the article: a researcher could provide a DNA sequence and ask whether a mutation might increase a person’s disease risk. Carbon-X would then provide a prediction and an explanation, according to Hugging Face. The model can also generate new DNA sequences, which the company says could help scientists develop new gene therapies.

Carbon-X is an open-source model created with research labs in Italy and China. Hugging Face built it by training Qwen3-14B on genomic sequences and natural language, using DNA sequences from humans, other mammals, plants, and fungi. Tunstall said one motivation was to give the open research community tools that do not make researchers fully dependent on a single external AI provider.

Researchers can run Carbon-X on their own computers, which could help them avoid sharing sensitive patient data with outside AI companies. The article says viruses were explicitly excluded from the model’s training data to reduce misuse risks. Tunstall framed Carbon-X as a small step toward AI-assisted personalized medicine, where patient records and genomes could eventually help identify or design treatments.