Hugging Face has released Carbon-X, an open-source AI model that lets researchers ask questions about DNA in plain English. Users paste in a genetic sequence, type a question and get an answer along with an explanation of the model’s reasoning.
What can Carbon-X do with DNA?
A researcher could ask whether a particular mutation is likely to be harmful, and the model will predict an outcome and explain why. It can also write new DNA sequences, a capability that could support work on gene therapies, Business Insider reported.
How was Carbon-X built?
The model reads genetic code and ordinary words in the same stream. Hugging Face built it on Alibaba’s Qwen3-14B language model, which has about 14.8 billion parameters. The company then trained it on 302 billion tokens of DNA, English text and biology material. The model card lists the Telethon Institute of Genetics and Medicine and the University of Naples Federico II in Italy, along with the Zhongguancun Academy in China, as partners.
Lewis Tunstall, a senior research scientist at Hugging Face, said the team left viruses out of the training data to reduce the risk of misuse. He said the project is meant to give open researchers tools they control, instead of forcing them to depend on large labs such as OpenAI and Anthropic.
“We wanted to provide the open research community with tools,” Tunstall told Business Insider.
The open release is a central selling point. Labs can run the model on their own hardware, which means sensitive patient data does not have to leave their systems. The weights fit on a single 80GB graphics card and are released under the Apache 2.0 license.
On Hugging Face’s own tests, Carbon-X averaged 66.7 percent. That places it just behind OpenAI’s GPT-6.1 Sol, which scored 67.5 percent, at a small fraction of the cost per task, according to the company. TNW has not independently verified those results.
Is Carbon-X ready for medical decisions?
The model card also carries a warning. Carbon-X has not been clinically validated, and it should not be used for medical decisions. Its predictions have not yet been tested in lab experiments, Wired reported.
Tunstall pointed to the longer-term goal of personalized medicine. “The holy grail is, in general, this concept of personalized medicine,” he told Business Insider, describing a future in which people could study their own medical records and genomes.
The release comes weeks after Nvidia agreed to buy Hugging Face for about $12.9 billion. The deal is expected to close in the first half of 2027.
Questo articolo è stato scritto con l'assistenza dell'IA.
News Factory APP - notizie agentiche per potenziare il tuo SEO e AEO.