Can Qwen 3.5 397B A17B run on NVIDIA GB200 192GB?

YES — With Q3_K_S

S86Excellent

○Estimated from fit model

Qwen 3.5 397B A17B needs ~217.5 GB VRAM. NVIDIA GB200 192GB has 192.0 GB. With Q3_K_S quantization, expect ~73 tok/s.

Runtime: llama.cppCapacity: OffloadBandwidth: HighStack: StandardBottleneck: Host offload

Operating mode

Choose the run profile you care about

Interactive favors responsiveness, while light API and scale-out lean harder on serving readiness. The fit stays the same, but the recommendation lens changes.

Current mode

Balanced

Balanced for general local use. Keeps the ranking neutral across personal and serving workflows.

Qwen 3.5 397B A17B at Q4_K_M needs 265.1 GB — too much for NVIDIA GB200 192GB (192.0 GB). Runs at Q3_K_S (217.5 GB) with low quality. 2 quantization levels fit.

Capabilities:

Select quantization to explore

Q4_K_M (Medium quality) — 265.1 GB, exceeds 192.0 GB available

265.1 GB required192.0 GB available

138% VRAM needed

73.1 GB over capacity — needs offload or smaller quantization

Fit status

Too heavy

Decode

45.5 tok/s

TTFT

4258 ms

Safe context

Memory

265.1 GB / 192.0 GB

Offload

30%

Memory breakdown

Weights242.2 GB

KV Cache2.9 GB

Runtime0.9 GB

Headroom19.2 GB

See how fast it feels

With memory offload — actual speed may be lower

See how fast it feelsQwen 3.5 397B A17B on NVIDIA GB200 192GB
1st promptCold start — includes initialization
>Write a Python function to sort a list
Here's a Python function that sorts a list using quicksort:

def quicksort(arr):
 if len(arr) <= 1:
 return arr
 pivot = arr[len(arr) // 2]
 left = [x for x in arr if x < pivot]
 middle = [x for x in arr if x == pivot]
 right = [x for x in arr if x > pivot]
 return quicksort(left) + middle + quicksort(right)

# Example usage
numbers = [3, 6, 8, 10, 1, 2, 1]
print(quicksort(numbers)) # [1, 1, 2, 3, 6, 8, 10]

The algorithm picks a pivot, partitions the list into smaller, equal, and larger elements, then recursively sorts each partition and concatenates the results.
2nd promptWarm — model ready, no init overhead
>Now add type hints and a docstring
def quicksort(arr: list[int]) -> list[int]:
 """Sort a list using the quicksort algorithm.

 Args:
 arr: List of integers to sort.

 Returns:
 A new sorted list.
 """
 if len(arr) <= 1:
 return arr
 pivot = arr[len(arr) // 2]
 left = [x for x in arr if x < pivot]
 middle = [x for x in arr if x == pivot]
 right = [x for x in arr if x > pivot]
 return quicksort(left) + middle + quicksort(right)

Added type hints for the input and return type, plus a Google-style docstring.
Estimated: 45.5 tok/s decode · 4.3s TTFT (warm) · 114 tok/s prefill

What limits this setup

It fits through host-memory offload, and offload is the main reason performance drops.

CPU or host-memory offload is active

About 10% of the working set spills out of accelerator memory, which usually hurts latency and sustained decode throughput.

Very little memory headroom

You can run the model, but there is not much room left for longer context, bigger batches, extra apps, or future model updates.

Best improvement path

Remove offload with more accelerator memory

Prioritize a GPU or unified-memory tier that fits the whole model natively. Removing offload usually helps more than small compute gains.

Buy headroom, not only minimum fit

A slightly larger memory tier gives you safer context growth and makes the recommendation more future-proof.

Increase host RAM if you keep offloading

This setup may need roughly 22.8 GB of extra host RAM just for the offloaded portion, before OS and other tools.

Performance by workload

Workload	Grade	Fit	Decode	TTFT	Context
Chat	F	Too heavy	45.9 tok/s	2302 ms	4K
Coding	F	Too heavy	45.5 tok/s	4258 ms	4K
Agentic Coding	F	Too heavy	44.7 tok/s	6304 ms	4K
Reasoning	F	Too heavy	45.5 tok/s	5032 ms	4K
RAG	F	Too heavy	44.7 tok/s	7880 ms	4K

Quantization options

How Qwen 3.5 397B A17B (397B params) fits at each quantization level on NVIDIA GB200 192GB (192.0 GB usable).

Quant	Bits	VRAM	Quality	Fit
Q2_K	2	154.8 GB	Low	F0
Q3_K_S	3	194.5 GB	Low	F0
NVFP4	4	222.3 GB	Medium	F0
Q4_K_M	4	242.2 GB	Medium	F0
Q5_K_M	5	285.8 GB	High	F0
Q6_K	6	325.5 GB	High	F0
Q8_0	8	424.8 GB	Very High	F0
F16	16	813.8 GB	Maximum	F0

Get started

Copy-paste commands to run Qwen 3.5 397B A17B on your machine.

Run

docker run --rm -it ghcr.io/ggerganov/llama.cpp:full \
 --hf-repo "Qwen/Qwen3.5-397B-A17B-Instruct" \
 --hf-file "Qwen3.5-397B-A17B-Instruct-Q4_K_M.gguf" \
 -c 4096 -ngl 99

Upgrade options