mlxradar
Alibaba Qwen · First MLX build 29 Jul 2025

Qwen3-30B-A3B-Instruct-2507

MLX Text apache-2.0

A 31B parameter text generation model from Alibaba Qwen, with 16 MLX builds, from 4-bit to bf16.

Hugging Face
Parameters
31B
From Hugging Face
Quantisation
4-bit
Qwen3-30B-A3B-Instruct-2507-4bit
Memory needed
22 GB
Estimate, unified memory
Download
17.2 GB
Weights, estimated
Downloads
6.5K
This build, last 30 days
Task
Text generation
Text
MLX builds
16
On mlx-community
Devices
Mac
Apple silicon

Memory

Estimated from the size of the weights, plus a fifth for context and working memory and 2 GB for macOS and your apps. Long contexts need more.

Run it

Terminal · mlx-lm
$ pip install -U mlx-lm
$ mlx_lm.generate --model mlx-community/Qwen3-30B-A3B-Instruct-2507-4bit --prompt "Hello"

MLX builds

16 on mlx-community
Click a column to sort. Smallest size and most downloads are highlighted.
Qwen3-30B-A3B-Instruct-2507-4bit 4-bit 17.2 GB 22 GB 6.5K
Qwen3-30B-A3B-Instruct-2507-5bit 5-bit 21.0 GB 26 GB 68
Qwen3-30B-A3B-Instruct-2507-5bit-DWQ-lr9e-8 5-bit DWQ 21.0 GB 26 GB 46
Qwen3-30B-A3B-Instruct-2507-5bit-DWQ-lr5e-7 5-bit DWQ 21.0 GB 26 GB 34
Qwen3-30B-A3B-Instruct-2507-5bit-DWQ-lr2e-7 5-bit DWQ 21.0 GB 26 GB 29
Qwen3-30B-A3B-Instruct-2507-5bit-DWQ-lr3e-8 5-bit DWQ 21.0 GB 26 GB 29
Qwen3-30B-A3B-Instruct-2507-5bit-DWQ-lr1e-7 5-bit DWQ 21.0 GB 26 GB 26
Qwen3-30B-A3B-Instruct-2507-6bit 6-bit 24.8 GB 30 GB 77
Qwen3-30B-A3B-Instruct-2507-6bit-DWQ-lr1e-8 6-bit DWQ 24.8 GB 30 GB 61
Qwen3-30B-A3B-Instruct-2507-6bit-DWQ-lr5e-8 6-bit DWQ 24.8 GB 30 GB 32
Qwen3-30B-A3B-Instruct-2507-6bit-DWQ-lr3e-7 6-bit DWQ 24.8 GB 30 GB 29
Qwen3-30B-A3B-Instruct-2507-6bit-DWQ-lr8e-7 6-bit DWQ 24.8 GB 30 GB 28
Qwen3-30B-A3B-Instruct-2507-6bit-DWQ-lr8e-8 6-bit DWQ 24.8 GB 30 GB 28
Qwen3-30B-A3B-Instruct-2507-6bit-DWQ-lr5e-9 6-bit DWQ 24.8 GB 30 GB 26
Qwen3-30B-A3B-Instruct-2507-8bit 8-bit 32.4 GB 39 GB 217
Qwen3-30B-A3B-Instruct-2507-bf16 bf16 61.1 GB 71 GB 39

Sizes are estimated from Hugging Face’s parameter counts and the quantisation. Downloads are for the last 30 days.

Similar models

Qwen3-Coder-30B-A3B-InstructTrending
Alibaba Qwen · 31B · apache-2.0
MLXCore AI
Quant3-bit / 4-bit … Memory17 GB Downloads27K Added31 Jul 2025
Qwen3-30B-A3B
Alibaba Qwen · 31B · apache-2.0
MLX
Quant3-bit / 4-bit … Memory17 GB Downloads10K Added28 Apr 2025
Qwen3-30B-A3B-Thinking-2507
Alibaba Qwen · 31B · apache-2.0
MLX
Quant4-bit / bf16 Memory22 GB Downloads253 Added30 Jul 2025
Qwen3-32B
Alibaba Qwen · 33B · apache-2.0
MLX
Quant3-bit / 4-bit … Memory19 GB Downloads2.4K Added28 Apr 2025

Changelog

  1. MLX 6-bit DWQ. Qwen3-30B-A3B-Instruct-2507-6bit-DWQ-lr5e-9 published on mlx-community.
  2. MLX 6-bit DWQ. Qwen3-30B-A3B-Instruct-2507-6bit-DWQ-lr1e-8 published on mlx-community.
  3. MLX 6-bit DWQ. Qwen3-30B-A3B-Instruct-2507-6bit-DWQ-lr8e-7 published on mlx-community.
  4. MLX bf16. Qwen3-30B-A3B-Instruct-2507-bf16 published on mlx-community.
  5. MLX 6-bit DWQ. Qwen3-30B-A3B-Instruct-2507-6bit-DWQ-lr8e-8 published on mlx-community.
  6. MLX 6-bit DWQ. Qwen3-30B-A3B-Instruct-2507-6bit-DWQ-lr5e-8 published on mlx-community.
  7. MLX 5-bit DWQ. Qwen3-30B-A3B-Instruct-2507-5bit-DWQ-lr1e-7 published on mlx-community.
  8. MLX 5-bit DWQ. Qwen3-30B-A3B-Instruct-2507-5bit-DWQ-lr9e-8 published on mlx-community.