Skip to content

clef:flash-fp16

3 TagsUpdated 9B params4096 contextEnglishApache-2.0by Cloudflare

Cloudflare's Clef-Flash: Qwen3.5-9B post-trained with a joint schema head that scores every option of every question together, in one forward pass per request. Needs about 19 GB of memory. This tag computes in fp16 on NVIDIA GPUs: faster than clef:flash, with at least 99.5 % of its decisions the same. On a CPU it runs, slowly; use clef:flash there.

9b
ollaya run clef:flash-fp16 --preset triage "I was charged twice for my subscription this month and want a refund."

Details

  • graph27f9da9d85c7 · 13 MBonnx · clef · 9B · fp16
  • weights8b45a8e96814 · 4.9 GBhuggingface.co/Cloudflare/clef-flash/resolve/17f0b0a…/model-00001-of-00004.safetensors
  • weights7590856c713e · 5.0 GBhuggingface.co/Cloudflare/clef-flash/resolve/17f0b0a…/model-00002-of-00004.safetensors
  • weightse6eac2467952 · 5.0 GBhuggingface.co/Cloudflare/clef-flash/resolve/17f0b0a…/model-00003-of-00004.safetensors
  • weights9fcecc6556b3 · 3.9 GBhuggingface.co/Cloudflare/clef-flash/resolve/17f0b0a…/model-00004-of-00004.safetensors
  • weights19cdcec8c81d · 244 MBhuggingface.co/Cloudflare/clef-flash/resolve/17f0b0a…/joint_head.safetensors
  • tokenizer06b9509352d2 · 20 MBhuggingface.co/Cloudflare/clef-flash/resolve/17f0b0a…/tokenizer.json
  • decision35a9f4935cff · 3 KB{"engine": "onnx", "family": "clef", "encoder": "", "layout": "clef-joint-v1", …}
  • calibration16180173d812 · 165 B{"temperature": [1.0, 1.0, 1.0]}
  • license0e73553d4bd8 · 10 KBClef-Flash by Cloudflare (https://huggingface.co/Cloudflare/clef-flash): Qwen3.5-9B, fully post-trained, with a joint schema head, Apache-2.0.

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.

Clef readme and all models