Skip to content

sura:2b

3 TagsUpdated 2B params8192 contextEnglishApache-2.0by Ollaya

Qwen3.5-2B: 0.570 on typed decisions, zero-shot; 63 ms for five questions on an RTX 5090.

2b
ollaya run sura:2b --preset triage "I was charged twice for my subscription this month and want a refund."

Details

  • graphdd513c73ce4e · 8 MBonnx · pplx · 2B · fp32
  • weightsd775fca764c0 · 4.4 GBhuggingface.co/ollaya-dev/sura-2b/resolve/9676c6b…/model.safetensors
  • weights503349bee928 · 1 MBhuggingface.co/ollaya-dev/sura-2b/resolve/9676c6b…/readout.safetensors
  • tokenizer6f32ce20dc35 · 20 MBhuggingface.co/ollaya-dev/sura-2b/resolve/9676c6b…/tokenizer.json
  • decision15b9c542ca22 · 6 KB{"engine": "onnx", "family": "pplx", "encoder": "", "layout": "pplx-joint-v1", …}
  • calibration2745b9c32f7e · 181 B{"temperature": [0.92, 0.92, 0.92]}
  • licensee883397b3998 · 11 KBSura by Ollaya (https://huggingface.co/ollaya-dev)

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.

sura readme and all models