{"version":1,"type":"story","url":"https://digestai.news/story/hugging-face-releases-carbon-x-a-14-8b-open-dna-language-model","json":"https://digestai.news/story/hugging-face-releases-carbon-x-a-14-8b-open-dna-language-model.json","markdown":"https://digestai.news/story/hugging-face-releases-carbon-x-a-14-8b-open-dna-language-model.md","slug":"hugging-face-releases-carbon-x-a-14-8b-open-dna-language-model","headline":"Hugging Face releases Carbon-X, a 14.8B open DNA language model","summary":"Hugging Face has released Carbon-X, an open-source DNA language model with 14.8B parameters. Built in collaboration with Zhongguancun Academy and TIGEM / University of Naples Federico II, the model is based on Qwen3-14B and uses a hybrid tokenizer to process raw DNA sequences and natural language within a single decoder. The weights are available under the Apache 2.0 license, allowing researchers to keep sensitive genetic data on local hardware.\n\nThe model was trained in four stages, including embedding warmup, mid-training on 302B tokens, supervised fine-tuning, and reinforcement learning. Hugging Face reports that Carbon-X achieves a 66.7% average across five DNA benchmarks, outperforming other open models. It scores 74.9% on BioReason Coding, leading all tested models, though it trails GPT-6.1 Sol by 0.8 points overall. Performance on splice-site questions remains low at 29.8%, and the model is not clinically validated.\n\nCarbon-X is designed for cost-effective inference, running on a single 80 GB GPU. Hugging Face estimates that scoring 2.4 million ClinVar variants would cost approximately $300 using Carbon-X, compared to $32,000 to $4 million for frontier models. The release includes deployment options via vLLM, Transformers, and llama.cpp, as well as agent tools for integration into existing workflows.","keyPoints":["Carbon-X is a 14.8B parameter open model released by Hugging Face under Apache 2.0 license.","It scores 66.7% on five DNA benchmarks, leading open models but trailing GPT-6.1 Sol by 0.8 points.","Hugging Face estimates scoring 2.4 million ClinVar variants costs about $300 on Carbon-X versus up to $4 million on frontier models."],"whyItMatters":"Carbon-X provides a low-cost, open-weight option for genomic analysis, enabling labs to process sensitive DNA data locally without relying on expensive proprietary APIs. This reduces barriers to entry for computational biology research.","category":{"slug":"models","name":"Generative AI & Models","url":"https://digestai.news/category/models"},"entities":{"companies":["Hugging Face","Zhongguancun Academy","TIGEM","University of Naples Federico II"],"models":["Carbon-X","Qwen3-14B","GPT-6.1 Sol","Opus 5.5","Gemini 4 Argon","Kimi K3"],"people":[]},"firstPublishedAt":"2026-10-25T16:53:40Z","updatedAt":"2026-10-25T16:53:40Z","sourceCount":1,"hasPrimarySource":false,"sources":[{"outlet":"MarkTechPost","title":"Hugging Face Releases Carbon-X, a 14B Open DNA Language Model that Brings DNA and Natural Language Together for the First Time","url":"https://marktechpost.com/2026/10/25/hugging-face-carbon-x-14b-dna-language-model","publishedAt":"2026-10-25T16:53:40Z","type":"press","primary":false,"lead":true}],"sourceNotes":null,"discussions":[],"thread":{"title":"AI Meets Genetics From Brain to DNA","url":"https://digestai.news/thread/researchers-introduce-seal-ai-model-for-gene-regulation-in-developing-human","storyCount":3},"cite":{"text":"Digest AI, \"Hugging Face releases Carbon-X, a 14.8B open DNA language model\", 25 October 2026, https://digestai.news/story/hugging-face-releases-carbon-x-a-14-8b-open-dna-language-model","publisher":"Digest AI","title":"Hugging Face releases Carbon-X, a 14.8B open DNA language model","datePublished":"2026-10-25T16:53:40Z","url":"https://digestai.news/story/hugging-face-releases-carbon-x-a-14-8b-open-dna-language-model"},"generatedBy":"Written by Digest AI's editorial model from the linked sources; the sources are the record.","license":"Headlines, digests and key points are written by Digest AI and may be quoted with a link to the story page. Linked articles belong to their publishers. Terms: https://digestai.news/terms#reuse"}