---
license: apache-2.0
tags:
- mlx
- distill
- cli
- code
- qwen
- quantized
- expert-model
- domain-specific
pipeline_tag: text-generation
base_model: samuelfaj/distill-1.7B-MLX
---

# distill-1.7B-4bit-MLX

4-bit quantized version of **[distill-1.7B](https://huggingface.co/samuelfaj/distill-1.7B-MLX)** — an Expert Language Model for CLI output compression. **Zero accuracy loss vs fp16.**

| Format | Size | Accuracy |
|--------|------|----------|
| fp16 | 3.2 GB | 95% |
| **4-bit** | **1.0 GB** | **95%** |

Built for [distill](https://github.com/samuelfaj/distill). [Full collection](https://huggingface.co/collections/samuelfaj/distill-6a0606f9b131c289025659fc)