Can DeepSeek V3.2 run on AMD Instinct MI325X 256GB?

Q: What should I upgrade first if DeepSeek V3.2 feels slow on AMD Instinct MI325X 256GB?

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.

YES — With Q2_K

A78Great

Estimated from fit model

DeepSeek V3.2 needs ~288.7 GB VRAM. AMD Instinct MI325X 256GB has 256.0 GB. With Q2_K quantization, expect ~27 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.

DeepSeek V3.2 at Q4_K_M needs 436.3 GB — too much for AMD Instinct MI325X 256GB (256.0 GB). Runs at Q2_K (288.7 GB) with low quality.

Capabilities:

Select quantization to explore

Q4_K_M (Medium quality) — 436.3 GB, exceeds 256.0 GB available

436.3 GB required256.0 GB available

170% VRAM needed

180.3 GB over capacity — needs offload or smaller quantization

Fit status

Too heavy

Decode

8.3 tok/s

TTFT

23195 ms

Safe context

Memory

436.3 GB / 256.0 GB

Offload

40%

Memory breakdown

Weights409.3 GB

KV Cache0.5 GB

Runtime0.9 GB

Headroom25.6 GB

See how fast it feels

With memory offload — actual speed may be lower

See how fast it feelsDeepSeek V3.2 on AMD Instinct MI325X 256GB
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: 8.3 tok/s decode · 23.2s TTFT (warm) · 21 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 29.6 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	8.4 tok/s	12638 ms	4K
Coding	F	Too heavy	8.3 tok/s	23195 ms	4K
Agentic Coding	F	Too heavy	8.3 tok/s	33814 ms	4K
Reasoning	F	Too heavy	8.3 tok/s	27412 ms	4K
RAG	F	Too heavy	8.3 tok/s	42267 ms	4K

Quantization options

How DeepSeek V3.2 (671B params) fits at each quantization level on AMD Instinct MI325X 256GB (256.0 GB usable).

Quant	Bits	VRAM	Quality	Fit
Q2_K	2	261.7 GB	Low	F0
Q3_K_S	3	328.8 GB	Low	F0
NVFP4	4

Get started

Copy-paste commands to run DeepSeek V3.2 on your machine.

Run

ollama run deepseek-v3.2

Frequently asked questions

Can AMD Instinct MI325X 256GB run DeepSeek V3.2?

Yes, AMD Instinct MI325X 256GB can run DeepSeek V3.2 at Q2_K quantization (Very compromised (needs ~29.6 GB host RAM)). The recommended Q4_K_M requires 436.3 GB which exceeds available memory, but at Q2_K it needs only 288.7 GB. Expected decode speed: 26.5 tok/s.

How much VRAM does DeepSeek V3.2 need?

DeepSeek V3.2 (671B parameters) requires approximately 436.3 GB at Q4_K_M quantization. On AMD Instinct MI325X 256GB, it fits at Q2_K using 288.7 GB.

What is the best quantization for DeepSeek V3.2?

The recommended quantization is Q4_K_M, but on AMD Instinct MI325X 256GB the best fitting quantization is Q2_K, which uses 288.7 GB.

What speed will DeepSeek V3.2 run at on AMD Instinct MI325X 256GB?

On AMD Instinct MI325X 256GB, DeepSeek V3.2 achieves approximately 26.5 tokens per second decode speed with a time-to-first-token of 7313ms using Q2_K quantization.

Can AMD Instinct MI325X 256GB run DeepSeek V3.2 for coding?

For coding workloads, DeepSeek V3.2 on AMD Instinct MI325X 256GB receives a F grade with 8.3 tok/s and 4K context.

What context window can DeepSeek V3.2 use on AMD Instinct MI325X 256GB?

On AMD Instinct MI325X 256GB, DeepSeek V3.2 can safely use up to 4K tokens of context at Q2_K quantization. The model's official context limit is 128K, but available memory constrains the safe maximum.

See all results for AMD Instinct MI325X 256GB See all hardware for DeepSeek V3.2

Embed this result▼

Paste this snippet into any page to show a live fit card.

<iframe src="https://willitrunai.com/embed/deepseek-v3.2-671b-on-instinct-mi325x-256gb" width="400" height="180" frameborder="0" style="border:none;border-radius:12px;overflow:hidden;" title="Will It Run AI — fit result"></iframe>

Preview: