Skip to content

d1-omni:600m

2 TagsUpdated 587M params16384 contextEnglish, German, Spanish, French, it, nl, pl, pt, ar, hi, ja, ru, Turkish, vi, ChineseLFM Open License v1.0by Liquid AI

d1-omni-600M (LFM2.5-Encoder-350M trunk; 587M in the checkpoint, 381M read for text), fp32: text questions in 15 languages, 16,384-token context.

587m
ollaya run d1-omni:600m --preset triage "I was charged twice for my subscription this month and want a refund."

Details

  • graph3c3db53090a3 · 4 MBonnx · d1-omni · 587M · fp32
  • weights0713bb05270c · 2.3 GBhuggingface.co/LiquidAI/d1-omni-600M/resolve/02b55d7…/model.safetensors
  • tokenizer1efc3a6609ab · 5 MBhuggingface.co/LiquidAI/d1-omni-600M/resolve/02b55d7…/tokenizer.json
  • decision7a0969de284a · 1 KB{"engine": "onnx", "family": "d1-omni", "encoder": "", "layout": "d1-omni-markers-v1", …}
  • calibration9db77c5aad75 · 443 B{"temperature_by_options": {"choice:11+": …, "choice:2": …, "choice:3-5": …, "choice:6-10": …, "noul:2": …, "score:3-5": …, "score:6-10": …}}
  • licensed31683c712de · 11 KBd1-omni-600M by Liquid AI (https://huggingface.co/LiquidAI/d1-omni-600M).

Every layer is checked against its sha256 when it is pulled. Weights and tokenizers download from the model author's Hugging Face repository at a pinned commit; Ollaya never re-hosts them.

d1-omni readme and all models