Alibaba
Qwen 3 235B A22B (235B parameters) requires approximately 147.7 GB of VRAM with Q4_K_M quantization. As a Mixture of Experts model with 22B active parameters, it uses less memory than its total parameter count suggests. For the best balance of quality and speed, we recommend hardware with at least 170 GB of VRAM.
Get started
— copy & paste to run locallyCopy-paste commands to run Qwen 3 235B A22B on your machine.
Run
lms load Qwen3-235B-A22B-Instruct-2507 && lms server startQuick specs
About this model
Related models
Quick picks
Best hardware
Run this model
Quantization options
No hardware detected — fit column shows raw VRAM estimates
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 91.7 GB | Low | — |
Q3_K_S | 3 | 115.2 GB | Low | — |
NVFP4 | 4 | 131.6 GB | Medium | — |
Q4_K_M | 4 | 143.4 GB | Medium | — |
Q5_K_M | 5 | 169.2 GB | High | — |
Q6_K | 6 | 192.7 GB | High | — |
Q8_0 | 8 | 251.5 GB | Very High | — |
F16 | 16 | 481.7 GB | Maximum | — |
Quality benchmarks
Coding
Reasoning
General
Source: official · 2025-07-14
Hardware compatibility
Computing compatibility...
Memory breakdown
Frequently asked questions
Qwen 3 235B A22B (235B parameters) requires approximately 147.7 GB of VRAM with Q4_K_M quantization. Lower quantizations like Q4_K_M use less memory but may reduce quality.
Yes, Mac Studio M3 Ultra 256GB can run Qwen 3 235B A22B with a compatibility score of 86/100. It provides 256 GB of memory and achieves approximately 11.3 tokens per second.
The recommended quantization for Qwen 3 235B A22B is Q4_K_M, which offers the best balance between model quality and memory efficiency. Higher quantizations preserve more quality but require more VRAM.
The top recommended hardware for Qwen 3 235B A22B: AMD Instinct MI325X 256GB (score: 95/100), AMD Instinct MI350X 288GB (score: 93/100), NVIDIA B200 180GB (score: 92/100). These provide the best combination of memory, bandwidth, and compute for running this model locally.
Yes, Qwen 3 235B A22B is well-suited for chat as well as reasoning. It was designed with these use cases in mind.
See also
