Run Qwen3 235B-A22B locally

Which workstation runs Qwen3 235B-A22B?

A 235B mixture-of-experts model with 22B active per token. At 4-bit the weights alone need about 140 GB: more than any 128 GB desktop machine, so this is the model that separates desktop systems from GB300 workstations.

Parameters235B · 22B active

Weights at 4-bit139 GB

Published formatBF16 (published checkpoint)

LicenceApache 2.0

ReleasedAlibaba Qwen · April 2025

Short answer

8 of 24 catalogued workstations pass the memory screen at 4-bit. The cheapest with a public price is the Exxact Valence DGX Station (≈ $95k). On the Exxact Valence DGX Station, memory bandwidth caps one request at ≤ 499 tok/s. No platform has a documented recipe in the Lab yet.

Every workstation, screened

Estimate · 8K context · 1 request

Machine	Memory	Needed at 4-bit	Speed ceiling	Highest precision that fits	Recipe	List price
Exxact Valence DGX StationNVIDIA GB300	252 GB	147 GB Fits	≤ 499 tok/s	4-bit	None yet	≈ $95k
MSI XpertStation WS300NVIDIA GB300	252 GB	147 GB Fits	≤ 499 tok/s	4-bit	None yet	≈ $109k
Dell Pro Max with GB300NVIDIA GB300	252 GB	147 GB Fits	≤ 499 tok/s	4-bit	None yet	≈ $175k
NVIDIA DGX StationNVIDIA GB300	252 GB	147 GB Fits	≤ 499 tok/s	4-bit	None yet	Price from the manufacturer
ASUS ExpertCenter Pro ET900N G3NVIDIA GB300	252 GB	147 GB Fits	≤ 499 tok/s	4-bit	None yet	Price from the manufacturer
HP ZGX FuryNVIDIA GB300	252 GB	147 GB Fits	≤ 499 tok/s	4-bit	None yet	Price from the manufacturer
GIGABYTE W775-V10-L01NVIDIA GB300	252 GB	147 GB Fits	≤ 499 tok/s	4-bit	None yet	Price from the manufacturer
Supermicro Super AI Station · ARS-511GDNVIDIA GB300	252 GB	147 GB Fits	≤ 499 tok/s	4-bit	None yet	Price from the manufacturer
Framework Desktop · Ryzen AI Max+AMD Ryzen AI Max	128 GB	151 GB Too large	–	None	None yet	≈ $3.4k
MINISFORUM MS-S1 MAXAMD Ryzen AI Max	128 GB	151 GB Too large	–	None	None yet	≈ $3.8k
NVIDIA DGX SparkNVIDIA GB10	128 GB	151 GB Too large	–	None	None yet	≈ $4.7k
ASUS Ascent GX10NVIDIA GB10	128 GB	151 GB Too large	–	None	None yet	≈ $5.3k
Acer Veriton GN100NVIDIA GB10	128 GB	151 GB Too large	–	None	None yet	≈ $5.5k
MSI EdgeXpertNVIDIA GB10	128 GB	151 GB Too large	–	None	None yet	≈ $6k
Dell Pro Max with GB10NVIDIA GB10	128 GB	151 GB Too large	–	None	None yet	≈ $8.2k
Apple Mac Studio · M3 UltraApple Silicon	96 GB	151 GB Too large	–	None	None yet	Price from the manufacturer
HP ZGX Nano G1nNVIDIA GB10	128 GB	151 GB Too large	–	None	None yet	Price from the manufacturer
Lenovo ThinkStation PGXNVIDIA GB10	128 GB	151 GB Too large	–	None	None yet	Price from the manufacturer
GIGABYTE AI TOP ATOMNVIDIA GB10	128 GB	151 GB Too large	–	None	None yet	Price from the manufacturer
AMD Ryzen AI HaloAMD Ryzen AI Max	128 GB	151 GB Too large	–	None	None yet	Price from the manufacturer
HP Z8 Fury G6iNVIDIA RTX PRO	96 GB	147 GB Too large	–	None	None yet	Price from the manufacturer
Lenovo ThinkStation P7NVIDIA RTX PRO	96 GB	147 GB Too large	–	None	None yet	Price from the manufacturer
HP Z2 Mini G1aAMD Ryzen AI Max	128 GB	151 GB Too large	–	None	None yet	Price from the manufacturer
Apple Mac Studio · M5 UltraApple Silicon	96 GB	151 GB Too large	–	None	None yet	Preorder · price from the manufacturer

Fit: weights at 4-bit plus KV cache for 8K tokens and a runtime reserve, against 90% of one memory tier. Speed ceiling: memory bandwidth divided by the bytes read per generated token. Neither is a measured result. With only 22B active parameters, compute and runtime overhead limit speed well before bandwidth on the fastest machines. Method · Change the context, users or precision in the finder

Questions people ask

How much memory does Qwen3 235B-A22B need?

About 139 GB for the weights at 4-bit, plus the KV cache and runtime reserve: roughly 147 GB in total at 8K context and one request. Longer context and more simultaneous users need more.

What is the cheapest workstation that runs Qwen3 235B-A22B?

In the Lab’s catalogue, the Exxact Valence DGX Station (≈ $95k, Supplier list price · United States · Sept 2026) passes the memory screen at 4-bit.

How fast does Qwen3 235B-A22B run locally?

The Lab has no measured run yet. The memory-bandwidth ceiling for one request is ≤ 499 tok/s on the Exxact Valence DGX Station; real single-stream speed lands below the ceiling.

Other models

Run gpt-oss-120b locally Run Llama 3.3 70B Instruct locally Run Qwen3 32B locally

Head to head

NVIDIA DGX Spark vs Apple Mac Studio · M3 Ultra NVIDIA DGX Spark vs Apple Mac Studio · M5 Ultra NVIDIA DGX Spark vs Framework Desktop · Ryzen AI Max+ NVIDIA DGX Spark vs MINISFORUM MS-S1 MAX NVIDIA DGX Spark vs HP Z2 Mini G1a NVIDIA DGX Spark vs AMD Ryzen AI Halo Framework Desktop · Ryzen AI Max+ vs Apple Mac Studio · M3 Ultra Framework Desktop · Ryzen AI Max+ vs Apple Mac Studio · M5 Ultra HP ZGX Nano G1n vs HP Z2 Mini G1a

Ran Qwen3 235B-A22B on one of these machines? Publish your run with the exact model revision, runtime version, context and tokens per second. Measured results replace the estimates on this page.