Hugging Face releases free AI that lets scientists "chat with DNA"

Hugging Face, an AI company, has released a new model called Carbon-X that lets scientists ask questions about DNA in everyday language. The company announced it Thursday. Lewis Tunstall, a senior research scientist there, calls it a "first of its kind" tool that lets researchers essentially "chat with DNA."

What it does

Carbon-X can both analyze and generate DNA sequences, and it pairs that with plain language. Tunstall gave an example. A researcher could provide a DNA sequence and ask whether a certain mutation might raise a person's risk of disease. The model would offer a prediction and an explanation. It can also generate new DNA sequences, which could help researchers develop new gene therapies.

Hugging Face built it by training an existing language model, Qwen3-14B, on genomic sequences and natural language. The DNA came from humans, other mammals, plants and fungi. Labs in Italy and China helped. Carbon-X is open source, which means it is shared publicly for others to use and build on.

Privacy and safety choices

• Researchers can run it on their own computers, so sensitive patient data does not have to go to outside AI companies.
• Viruses were left out of the training data on purpose to avoid misuse.
• Tunstall says general chatbots sometimes refuse real biology questions, mistaking them for attempts to "design a bioweapon."

Why Hugging Face built it

Tunstall said the company wanted to keep biology researchers from depending on a few big AI labs. He pointed to mathematics, where he said pushing the frontier likely means working at a lab like OpenAI or Anthropic. He expects other sciences could follow. "We wanted to provide the open research community with tools that don't make them fully dependent on some external single provider," he said.

Hugging Face is a platform for sharing and developing AI models, and it also releases its own. Nvidia agreed in September to acquire it for nearly $13 billion. Tunstall also said troubling behavior seen in some larger AI models has not yet been observed in smaller, more specialized ones.

What it means for everyday people

Carbon-X is meant for researchers, not for the public. Tunstall described it as a small step toward personalized medicine, where people could use their own patient records and genomes to help find or even design their own treatments. He called that the "holy grail."

Plenty is still unknown. The material does not say how accurate the model's predictions are, how closely it has been tested, or when any treatment might come from it.