How much VRAM does Qwen 3 235B A22B need?

Qwen 3 235B A22B (235B parameters) requires approximately 161.8 GB of memory with Q4_K_M quantization.

What is the best quantization for Qwen 3 235B A22B?

The recommended quantization for Qwen 3 235B A22B is Q4_K_M, which balances quality and memory efficiency.

Can it run?

Yes, NVIDIA H200 PCIe 141GB can run Qwen 3 235B A22B with a C grade (Very compromised). Expected decode speed: 57.3 tok/s.

CUsable

Very compromised

Using Q4_K_M in llama.cpp

Capabilities:

Fit status

Very compromised

Decode

57.3 tok/s

TTFT

3380 ms

Safe context

14K

Memory

161.8 GB / 141.0 GB

Offload

10%

Weights143.4 GB

KV Cache3.4 GB

Runtime0.9 GB

Headroom14.1 GB

Workload	Grade	Fit	Decode	TTFT	Context
Agentic Coding	C	Very compromised (needs ~21 GB host RAM)	66.2 tok/s	4252 ms	27K
Chat	C	Very compromised (needs ~17.1 GB host RAM)	67.9 tok/s	1556 ms	7K
Coding	C	Very compromised	57.3 tok/s	3380 ms	14K
RAG	C	Very compromised (needs ~21 GB host RAM)	66.2 tok/s	5315 ms	27K
Reasoning	C	Very compromised (needs ~18.4 GB host RAM)	67.3 tok/s

How Qwen 3 235B A22B (235B params) fits at each quantization level on NVIDIA H200 PCIe 141GB (141.0 GB usable).

Quant	Bits	VRAM	Quality	Fit
Q2_KBest for your GPU	2	91.7 GB	Low	C43
Q3_K_S	3	115.2 GB	Low	C45

HuggingFace

huggingface-cli download qwen-3-235b-a22b

Upgrade options

AMD Instinct MI350X 288GBBudget pick

B108.7 tok/s decode

~$8,000 MSRP

AMD Instinct MI300X 192GBBest value

C76.9 tok/s decode

~$15,000 MSRP