AI Briefing
KOSign in

Carbon-X: 14B Open DNA Language Model Released with Local GGUF Support

·2026.10.22 19:28

Key point

HuggingFaceBio released Carbon-X, a 14B model midtrained on Qwen3 14B that outperforms open competitors on five DNA benchmarks and supports local inference via GGUF.

Details

  • HuggingFaceBio released Carbon-X, a 14B parameter model that integrates DNA and natural language processing, available on Hugging Face with a GGUF version for local execution via llama.cpp.
  • The model beats every open language model tested on five DNA benchmarks and matches GPT-6.1 Sol performance at approximately 1,000x lower cost.
  • Built by midtraining Qwen3 14B on a combination of DNA and natural language data using a new hybrid tokenizer, with a published recipe for the midtraining and RL process.
  • Use cases include querying DNA sequences in plain English, predicting mutation effects, and generating new DNA sequences.
  • The release aims to address data sensitivity and frontier model refusals in routine genomics, enabling legitimate research with open models.
open-sourcelocal-llmqwen3genomicscarbon-xdna-language-modelhuggingfacebio
Read originalShare on Threads

This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.

Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.

Related

RedditNace.ai releases Kapitsa, an open-source Apple Silicon inference engine for Qwen and NemotronRedditLLMxRay Released: Open-Source Observability Tool for Local Ollama ModelsHuggingFace BlogIBM releases Granite Speech 5.0 with improved ASR accuracy and 1.6x faster Turbo CTCHuggingFace BlogNVIDIA Releases Nemotron 3.5 Embed, Rerank, and Parse Models for Agent RetrievalGeekNewsCopilotKit Releases OpenIntelligentUI, an Open-Source Alternative to ChatGPT Intelligent UIKakaoKakao Releases Kanana-2-vision-encoder, Open-Sourcing Medium and Tiny Weights