Thursday, 22 October 2026 · Europe
EUR/USD 1.123 EUR/GBP 0.8383 EUR/CHF 0.9256 EUR/PLN 4.364 All rates →
Sign in · Join
EUROPES The European Report
European Edition Thursday, 22 October 2026
LATEST French weekend protests to be limited to union demos only, says Macron Le Flash →
Tech & Startups

Hugging Face releases Carbon-X, an AI model for questions about DNA

Hugging Face releases Carbon-X, an AI model for questions about DNA

Hugging Face has released Carbon-X, an open-source AI model that lets researchers ask questions about DNA in everyday language. Users paste in a genetic sequence, ask a question, and get an answer with an explanation. The model card shows its partners are the Telethon Institute of Genetics and Medicine and the University of Naples Federico […] This story continues at The Next Web

Hugging Face has released Carbon-X, an open-source AI model that lets researchers ask questions about DNA in everyday language. Users paste in a genetic sequence, ask a question, and get an answer with an explanation.

The model card shows its partners are the Telethon Institute of Genetics and Medicine and the University of Naples Federico II in Italy, and the Zhongguancun Academy in China.

Carbon-X starts from Alibaba’s Qwen3-14B language model and has about 14.8 billion parameters. Hugging Face then trained it on 302 billion tokens of DNA, English text and biology material, so the model reads genetic code and words in the same stream.

A researcher can, for example, ask whether a mutation is likely to be harmful. The model predicts and explains its reasoning. It can also write new DNA sequences, which could help work on gene therapies, Business Insider reported.

Lewis Tunstall, a senior research scientist at Hugging Face, told Business Insider that the team left viruses out of the training data to reduce the risk of misuse.

He also said the project aims to give open researchers tools they control, rather than leaving them to rely on large labs such as OpenAI and Anthropic.

“We wanted to provide the open research community with tools,” Tunstall told Business Insider.

Because the model is open, labs can run it on their own machines. As a result, 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 licence.

On Hugging Face’s own tests, Carbon-X scored an average of 66.7%. That puts it just behind OpenAI’s GPT-6.1 Sol at 67.5%, at a small fraction of the cost per task, the company said. TNW has not independently verified these results.

However, the model card states that Carbon-X has not been clinically validated and should not be used for medical decisions. Its predictions have also not yet been tested in lab experiments, Wired reported.

Tunstall told Business Insider that “the holy grail is, in general, this concept of personalised medicine,” where people could study their own medical records and genomes.

Hugging Face launched the model weeks after Nvidia agreed to buy the company for about $12.9bn. The deal is expected to close in the first half of 2027.

More from Tech & Startups