Irfanuruchi/Qwen2.5-0.5B-Instruct-MLX-4bit

Qwen2.5-0.5B-Instruct (MLX, 4-bit)

This repository contains an MLX-converted and 4-bit quantized version of Qwen/Qwen2.5-0.5B-Instruct.

• No fine-tuning or training was performed
• Format conversion + post-training quantization only
• Recommended default for on-device usage

Usage

pip install -U mlx-lm

mlx_lm.generate \
  --model Irfanuruchi/Qwen2.5-0.5B-Instruct-MLX-4bit \
  --prompt "Write a helpful onboarding message for an iOS app in 3 bullet points."

Bench notes (MacBook Pro M3 Pro)

• Prompt tokens: 45
• Generation tokens: 100
• Generation speed: ~292.9 tokens/sec
• Peak memory: ~0.319 GB

Tooling

• mlx-lm: 0.30.2
• mlx: bundled with Apple MLX (no public version string)

Related models

• 8-bit variant (higher quality):
  https://huggingface.co/Irfanuruchi/Qwen2.5-0.5B-Instruct-MLX-8bit