Attache Router 0.8B (MLX 4-bit)
Qwen3.5-0.8B fine-tuned (LoRA, merged) to turn one sentence said to an offline phone secretary into one JSON action
(create_event, query_events, edit_event, create_reminder, query_reminders, edit_reminder, search_contact, answer, clarify).
Text-only (vision tower removed), MLX affine 4-bit, group size 64. Trained only on synthetic sentences in Chinese, English and Spanish.
It expects the app's own system prompt and few-shot prefix; it is not a general chat model.
- Downloads last month
- -
Model size
0.8B params
Tensor type
U32
路
BF16 路
F32 路
Hardware compatibility
Log In to add your hardware
4-bit
Inference Providers NEW
This model isn't deployed by any Inference Provider. 馃檵 Ask for provider support