Science & Tech · AI · 2 hrs ago
Carbon-X lets users ask questions about DNA in everyday language
Hugging Face and academic partners released Carbon-X, an open-source AI model that connects DNA sequences with everyday language.
The project involves researchers at Zhongguancun Academy in China and the University of Naples Federico II in Italy.
Users can ask it questions about small sections of DNA and get conversational answers about mutations, gene activity and health.
DNA holds biological instructions in four-letter sequences, but scientists still do not understand how much of the human genome works.
The model may help researchers assess genetic changes as they study rare diseases and develop drugs or genetic therapies.
Its small size is designed to let it run on a user's own hardware, keeping sensitive genetic data on the user's device rather than sending it to a cloud service.
The developers say the model has not yet been validated, so its answers should not be treated as established medical guidance.
Hugging Face released Carbon-X, an open-source AI model that lets users ask questions about DNA in everyday language.
The model can help analyze genetic mutations, identify DNA regions involved in gene regulation and generate genomic sequences.
It was developed with researchers from Zhongguancun Academy in China and the University of Naples Federico II in Italy.
Carbon-X has 14 billion parameters and is designed to run on users’ own hardware, keeping genomic data off company servers.
Users can query DNA sequences of roughly 50,000 base pairs, but cannot upload full human genomes.
- Who
- Hugging Face and research collaborators at Zhongguancun Academy and the University of Naples Federico II developed Carbon-X.
- What
- They released an open-source AI model that can answer natural-language questions about DNA and perform genomic tasks.
- When
- October 22, 2026. Wired reported that it was released on Thursday.
- Where
- The model can run on users’ own hardware; the release location is not stated.
- Why
- It is intended to help researchers interpret genetic variants and support work on drugs, genetic therapies and precision medicine.
This story does not have two clearly opposing sides.
The team trained Carbon-X on 300 billion tokens combining DNA sequences, English text and examples linking DNA to biological descriptions.
Hugging Face released Carbon-X as an open-source model.
- Model
- Carbon-X
- Model size
- 14 billion parameters
- Training
- 300 billion tokens
- Query length
- Roughly 50,000 base pairs
- Release date
- October 22, 2026
- Collaborators
- Zhongguancun Academy and the University of Naples Federico II
3 sources
What people are saying
We've unlocked a new capability in language models, which is fundamentally about being able to chat with DNA
can actually ask questions about DNA that wouldn't be accessible through any other tool.