How much VRAM does Qwen3.5 397B A17B need?

Qwen3.5 397B A17B (397B parameters) requires approximately 331.0 GB of memory with Q4_K_M quantization.

What is the best quantization for Qwen3.5 397B A17B?

The recommended quantization for Qwen3.5 397B A17B is Q4_K_M, which balances quality and memory efficiency.

Can it run?

No, Qwen3.5 397B A17B requires more memory than AMD Instinct MI325X 256GB provides.

FWon't run

Too heavy

Using Q4_K_M in Ollama

Capabilities:

Fit status

Too heavy

Decode

17.8 tok/s

TTFT

10853 ms

Safe context

12K

Memory

331.0 GB / 256.0 GB

Weights242.2 GB

KV Cache62.0 GB

Runtime1.2 GB

Headroom25.6 GB

Workload	Grade	Fit	Decode	TTFT	Context
Agentic Coding	F	Too heavy	18.1 tok/s	15570 ms	21K
Chat	D	Very compromised (needs ~35.5 GB host RAM)	15.8 tok/s	6670 ms	7K
Coding	F	Too heavy	17.8 tok/s	10853 ms	12K
RAG	F	Too heavy	18.1 tok/s	19462 ms	21K
Reasoning	F	Too heavy	18.1 tok/s	12650 ms	12K

How Qwen3.5 397B A17B (397B params) fits at each quantization level on AMD Instinct MI325X 256GB (256.0 GB usable).

Quant	Bits	VRAM	Quality	Fit
Q2_K	2	154.8 GB	Low	C42
Q3_K_SBest for your GPU	3	194.5 GB	Low	C44

Upgrade options

AMD Instinct MI350X 288GBBest value

D21.3 tok/s decode

~$8,000 MSRP