Downloads · 30 days
24
43% of all-time downloads
adnank9/qwen3-1.7b-phone-assistant-dpo-experimental-4bit
qwen3-1.7b-phone-assistant-dpo-experimental-4bit is a machine learning model from adnank9. Use it for the machine learning task on the model card, and read the license before you ship it in a product. It is set up for mlx. The card lists the license as apache-2.0.
Experimental on-device model for Assistant, an iPhone app that answers calls and takes messages. DPO (QLoRA) directly on Qwen3-1.7B: teacher replies (Qwen3-8B) preferred over the model's own worst replies across 140+…
Downloads · 30 days
24
43% of all-time downloads
All-time downloads
56
Public
Parameters
1.7B
980 MB on disk
Likes
0
Public
Click a slice to open those files.
.safetensors968 MB · 99%
How the weights are stored.
U321.7B · 100%
From the Hugging Face model README
Experimental on-device model for Assistant, an iPhone app that answers calls and takes messages. DPO (QLoRA) directly on Qwen3-1.7B: teacher replies (Qwen3-8B) preferred over the model's own worst replies across 140+ generated caller personas.
Held-out evaluation, 32 unseen callers, same prompt and repeat guard:
| base | this model | |
|---|---|---|
| judge score /10 | 9.62 | 9.75 |
| repeated replies | 2.1% | 0% |
| words per reply | 21.6 | 20.4 |
| ended the call properly | 96.9% | 90.6% |
Not the app's default: it ends calls properly less often. Selectable in the app under Settings → Model.