Hugging Face Releases Carbon-X, an Open-Source AI Model That Answers Questions About DNA
Carbon-X runs locally on a laptop, keeping sensitive genomic data off Hugging Face and third-party servers. The 14-billion-parameter model was trained on 300 billion tokens and DNA grouped in six-letter sequences, with viruses excluded from its data.
Hugging Face released Carbon-X on Thursday, October 22, 2026 — an open-source AI model that allows users to ask conversational questions about DNA sequences and receive plain-language answers about how genetic changes may affect gene activity and health.
The model was developed in collaboration with researchers at Zhongguancun Academy in China and the University of Naples Federico II in Italy. Lewis Tunstall, a senior research scientist at Hugging Face, said Carbon-X is designed to help researchers develop new drugs and genetic therapies, and could potentially help patients interpret their own genomic data. “We’ve unlocked a new capability in language models, which is fundamentally about being able to chat with DNA,” Tunstall said.
Carbon-X was built by expanding an existing language model’s vocabulary to read DNA in six-letter groupings, then training it on 300 billion tokens combining DNA sequences, English text, and examples linking DNA to biological descriptions. The training data drew from mammals, vertebrates, invertebrates, plants, fungi, and protozoa. The team excluded viruses from training data to limit potentially dangerous applications.
The model contains 14 billion parameters and is small enough to run locally on a laptop, meaning no data is transmitted to Hugging Face or any third party — a feature Tunstall said was intentional given researchers’ restrictions around uploading sensitive patient data to cloud-based AI systems. Users can query DNA sequences of up to roughly 50,000 base pairs at a time, but cannot upload full human genomes.
One practical application is helping scientists determine whether a genetic mutation is benign or disease-causing — a critical step before developing targeted treatments. The developers caution, however, that Carbon-X has not yet been validated through wet lab experiments, so the accuracy of its predictions remains unconfirmed.
Hugging Face, which Nvidia agreed to acquire for nearly $13 billion in September, hosts a broad library of open-source and open-weight AI tools for developers, researchers, and businesses.