Skip to content

nimble:9b-fp16

4 TagsUpdated 9B params8192 contextEnglishApache-2.0by Bespoke Labs

Bespoke Labs' Nimble v2: a LoRA on Qwen3.5-9B that answers every question from the next-token logits of its option codes, calibrated with the author's temperature. Up to 255 options. Needs about 18 GB of memory; best on a 24 GB GPU. This tag computes in fp16 on NVIDIA GPUs: faster than nimble:9b, with at least 99.5 % of its decisions the same. On a CPU it runs, slowly; use nimble:9b there.

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

Details

  • graphb916c3c9595a · 15 MBonnx · nimble · 9B · fp16
  • weightsdb6f444b43d3 · 5.3 GBhuggingface.co/Qwen/Qwen3.5-9B/resolve/c202236…/model.safetensors-00001-of-00004.safetensors
  • weights31c7d7e2dd5d · 5.3 GBhuggingface.co/Qwen/Qwen3.5-9B/resolve/c202236…/model.safetensors-00002-of-00004.safetensors
  • weights7ec36ba3a417 · 5.4 GBhuggingface.co/Qwen/Qwen3.5-9B/resolve/c202236…/model.safetensors-00003-of-00004.safetensors
  • weightsb62b0c4cd7e4 · 3.3 GBhuggingface.co/Qwen/Qwen3.5-9B/resolve/c202236…/model.safetensors-00004-of-00004.safetensors
  • weights1bd126be997b · 173 MBhuggingface.co/bespokelabs/Bespoke-Nimble-9B-v2/resolve/4b8c04d…/adapter_model.safetensors
  • tokenizer06b9509352d2 · 20 MBhuggingface.co/bespokelabs/Bespoke-Nimble-9B-v2/resolve/4b8c04d…/tokenizer.json
  • decision24d679f54899 · 11 KB{"engine": "onnx", "family": "nimble", "encoder": "", "layout": "nimble-codes-v1", …}
  • calibration185dcc039a5e · 306 B{"temperature": [2.179, 2.179, 2.179]}
  • licenseff2de1e5553f · 10 KBBespoke-Nimble-9B-v2 by Bespoke Labs (https://huggingface.co/bespokelabs/Bespoke-Nimble-9B-v2)

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.

Nimble readme and all models