Can Qwen3-Coder 30B A3B Instruct run on AMD Instinct MI250 128GB?

YES — Runs Great

S91Excellent

Estimated from fit model

Qwen3-Coder 30B A3B Instruct needs ~33.8 GB VRAM. AMD Instinct MI250 128GB has 128.0 GB. With Q4_K_M quantization, expect ~329 tok/s.

Runtime: llama.cppCapacity: RoomyBandwidth: HighStack: StandardBottleneck: Balanced

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.

Capabilities:

Select quantization to explore

Q4_K_M (Medium quality) — 33.8 GB, 329.0 tok/s, Runs well

33.8 GB required128.0 GB available

26% VRAM used

Fit status

Runs well

Decode

329.0 tok/s

TTFT

588 ms

Safe context

256K

Memory

33.8 GB / 128.0 GB

Memory breakdown

Weights18.6 GB

KV Cache1.5 GB

Runtime0.9 GB

Headroom12.8 GB

See how fast it feels

See how fast it feelsQwen3-Coder 30B A3B Instruct on AMD Instinct MI250 128GB
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: 329.0 tok/s decode · 588ms TTFT (warm) · 823 tok/s prefill

What limits this setup

This setup is broadly balanced for this model.

No major red flags

This recommendation has enough memory headroom and acceptable estimated speed for the selected workload.

Best improvement path

Performance by workload

Workload	Grade	Fit	Decode	TTFT	Context
Chat	S	Runs well	329.0 tok/s	350 ms	256K
Coding	S	Runs well	329.0 tok/s	588 ms	256K
Agentic Coding	S	Runs well	329.0 tok/s	856 ms	256K
Reasoning	S	Runs well	329.0 tok/s	695 ms	256K
RAG	S	Runs well	329.0 tok/s	1070 ms	256K

Inference speed

Qwen3-Coder 30B A3B Instruct inference speed — tokens per second by GPU & Mac

Estimated decode speed (tokens/sec) for Qwen3-Coder 30B A3B Instruct at Q4_K_M across popular GPUs and Apple Silicon, using the fastest local runtime per device. Fastest is RTX 5090 32GB at ~182 tok/s. Speed is memory-bandwidth bound, so cards that fit the whole model in VRAM run far faster than ones that offload to system RAM.

GPU / Mac	Memory	Quant	Speed (tok/s)	Fits?
RTX 5090 32GB	32 GB	Q4_K_M	181.6	Fits
RTX 4090 24GB	24 GB	Q4_K_M	115.8	Offloads
RX 7900 XTX 24GB	24 GB	Q4_K_M	104.5	Offloads
RTX 3090 24GB	24 GB	Q4_K_M	99.1	Offloads
Mac Studio M3 Ultra 256GB	256 GB	Q4_K_M	84.2	Fits
Mac Studio M2 Ultra 128GB	128 GB	Q4_K_M	70.2	Fits
Mac Studio M1 Ultra 128GB	128 GB	Q4_K_M	66.5	Fits
MacBook Pro M4 Max 128GB	128 GB	Q4_K_M	52.0	Fits
MacBook Pro M4 Max 64GB	64 GB	Q4_K_M	52.0	Fits
MacBook Pro M3 Max 64GB	64 GB	Q4_K_M	36.3	Fits
MacBook Pro M1 Max 64GB	64 GB	Q4_K_M	33.3	Fits
RTX 4080 Super 16GB	16 GB	Q4_K_M	32.7	Too big
MacBook Pro M4 Pro 48GB	48 GB	Q4_K_M	31.8	Fits
RTX 4070 12GB	12 GB	Q4_K_M	11.4	Too big
RTX 3060 12GB	12 GB	Q4_K_M	7.2	Too big
RTX 4060 8GB	8 GB	Q4_K_M	4.5	Too big

Estimates for single-stream decoding at Q4_K_M; real tokens/sec varies with prompt length, context, batch size, and runtime build. Prompt processing (prefill) is faster than the decode figures shown here.

Quantization options

How Qwen3-Coder 30B A3B Instruct (30.5B params) fits at each quantization level on AMD Instinct MI250 128GB (128.0 GB usable).

Quant	Bits	VRAM	Quality	Fit
Q2_K	2	11.9 GB	Low	A82
Q3_K_S	3	14.9 GB	Low	A82
NVFP4	4	17.1 GB	Medium	A82
Q4_K_M	4	18.6 GB	Medium	A82
Q5_K_M	5	22.0 GB	High	A82
Q6_K	6	25.0 GB	High	A83
Q8_0	8	32.6 GB	Very High	A84
F16Best for your GPU	16	62.5 GB	Maximum	S89

Get started

Copy-paste commands to run Qwen3-Coder 30B A3B Instruct on your machine.

Run

ollama run qwen3-coder

Your hardware

More models your AMD Instinct MI250 128GB can run

Model	Params	Grade	Decode	Capabilities
Devstral 2 123B Instruct	123B	S	31.5 tok/s

Frequently asked questions

Can AMD Instinct MI250 128GB run Qwen3-Coder 30B A3B Instruct?

Yes, AMD Instinct MI250 128GB can run Qwen3-Coder 30B A3B Instruct with a S grade (Runs well). Expected decode speed: 329.0 tok/s.

How much VRAM does Qwen3-Coder 30B A3B Instruct need?

Qwen3-Coder 30B A3B Instruct (30.5B parameters) requires approximately 33.8 GB of memory with Q4_K_M quantization.

What is the best quantization for Qwen3-Coder 30B A3B Instruct?

The recommended quantization for Qwen3-Coder 30B A3B Instruct is Q4_K_M, which balances quality and memory efficiency.

What speed will Qwen3-Coder 30B A3B Instruct run at on AMD Instinct MI250 128GB?

On AMD Instinct MI250 128GB, Qwen3-Coder 30B A3B Instruct achieves approximately 329.0 tokens per second decode speed with a time-to-first-token of 588ms using Q4_K_M quantization.

Can AMD Instinct MI250 128GB run Qwen3-Coder 30B A3B Instruct for coding?

For coding workloads, Qwen3-Coder 30B A3B Instruct on AMD Instinct MI250 128GB receives a S grade with 329.0 tok/s and 256K context.

What context window can Qwen3-Coder 30B A3B Instruct use on AMD Instinct MI250 128GB?

On AMD Instinct MI250 128GB, Qwen3-Coder 30B A3B Instruct can safely use up to 256K tokens of context. The model's official context limit is 256K, but available memory constrains the safe maximum.

See all results for AMD Instinct MI250 128GB See all hardware for Qwen3-Coder 30B A3B Instruct

Embed this result▼

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

<iframe src="https://willitrunai.com/embed/qwen-3-coder-30b-a3b-on-instinct-mi250-128gb" width="400" height="180" frameborder="0" style="border:none;border-radius:12px;overflow:hidden;" title="Will It Run AI — fit result"></iframe>

Preview: