Hugging Face releases open Carbon-X model for DNA and natural language

Hugging Face has released Carbon-X, an open 14.8 billion parameter DNA language model, MarkTechPost reported. Hugging Face Bio developed the model in collaboration with the Zhongguancun Academy and TIGEM. It is a decoder-only model built on top of Qwen3-14B that reads raw DNA sequences within a chat format and returns answers in English. [1]

The model performs several biological analysis tasks, according to MarkTechPost. It can predict if genetic mutations are benign or pathogenic, identify regulatory regions, classify organisms, and predict the next bases in a sequence. Across five standard DNA benchmarks, Carbon-X achieved an average score of 66.7 percent. [1]

Large language models are machine learning systems trained on large datasets to analyze and generate text, providing the foundation for tools such as Qwen, Wikipedia notes. Hugging Face has released the Carbon-X weights under an Apache 2.0 open-source license, MarkTechPost reported. This distribution model allows research laboratories to run the system locally on their own graphics processing units to keep sensitive sequence data private. [1][3]