Hugging Face released Carbon-X, an open-source 14.8B parameter Apache 2.0 decoder-only model trained jointly on DNA and text that fits on an 80 GB GPU.

{
  "signal_id": "dc0ee20b-90f9-5897-a0e3-0f1b35982e6a",
  "signal_type": "reddit-company",
  "signal_subtype": "productLaunch",
  "detected_at": "2026-10-22T12:32:56+00:00",
  "company": {
    "name": "Hugging Face",
    "domain": "huggingface.co"
  },
  "data": {
    "nsfw": false,
    "stage": "confirmed",
    "awards": 0,
    "timing": "completed",
    "topics": [
      "ai",
      "open source",
      "genomics",
      "machine learning"
    ],
    "post_id": "1xchayq",
    "summary": "Hugging Face released Carbon-X, an open-source 14.8B parameter Apache 2.0 decoder-only model trained jointly on DNA and text that fits on an 80 GB GPU.",
    "evidence": [
      "[post] Carbon-X is an open 14.8B model trained jointly on DNA and text [R]",
      "[post] Hugging Face reports a 66.7% average across five DNA benchmarks with thinking disabled, versus 67.5% for GPT-6.1 Sol. The weights fit on one 80 GB GPU in bfloat16 and are Apache 2.0."
    ],
    "virality": "low",
    "post_date": "2026-10-22T12:32:56.000Z",
    "post_kind": "multi_media",
    "post_text": "Carbon-X is a 14.8B parameter decoder-only model that takes raw DNA in a normal chat and answers in English. It shares one token stream for text and DNA, using 6-mer tokens, and was mid-trained on 302B tokens.\n\nHugging Face reports a 66.7% average across five DNA benchmarks with thinking disabled, versus 67.5% for GPT-6.1 Sol. The weights fit on one 80 GB GPU in bfloat16 and are Apache 2.0. It has not been clinically validated, and the model card says exact sequence operations should still be handled by code.\n\nthe useful part is making sequence analysis local enough for sensitive datasets. has anyone tested whether the benchmark gains hold on private variant datasets?\n\nsource: https://huggingface.co/HuggingFaceBio/Carbon-X",
    "sentiment": "positive",
    "subreddit": "MachineLearning",
    "event_date": "2026-10",
    "post_flair": [
      "Research"
    ],
    "post_title": "Carbon-X is an open 14.8B model trained jointly on DNA and text [R]",
    "prominence": "core",
    "source_url": "https://www.reddit.com/r/MachineLearning/comments/1xchayq/carbonx_is_an_open_148b_model_trained_jointly_on/",
    "entity_role": "subject",
    "post_author": "jatingargiitk",
    "upvote_ratio": 0.9074074074074074,
    "mention_count": 5,
    "mention_surge": true,
    "subreddit_url": "https://www.reddit.com/r/MachineLearning/",
    "total_upvotes": 44,
    "comments_total": 1,
    "total_comments": 1,
    "other_companies": [
      {
        "name": "OpenAI",
        "role": "competitor",
        "domain": "openai.com"
      }
    ],
    "post_author_url": "https://www.reddit.com/user/jatingargiitk/",
    "signal_category": "event",
    "comments_included": 0,
    "products_mentioned": [
      "Carbon-X"
    ]
  }
}

curl https://signals.autobound.ai/v1/signals/dc0ee20b-90f9-5897-a0e3-0f1b35982e6a \
  -H "X-API-KEY: $AUTOBOUND_API_KEY"

curl -X POST https://signals.autobound.ai/v1/companies/enrich \
  -H "X-API-KEY: $AUTOBOUND_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"domain":"huggingface.co","limit":20}'