Can Llama 4 Maverick 17B 128E run on AMD Instinct MI325X 256GB?

YES — With Offload

A75Great

Estimated from fit model

Llama 4 Maverick 17B 128E needs ~273.4 GB VRAM. AMD Instinct MI325X 256GB has 256.0 GB. With Q4_K_M quantization, expect ~38 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.

Capabilities:

Select quantization to explore

Q4_K_M (Medium quality) — 273.4 GB, 37.9 tok/s, Runs with offload (needs ~15.6 GB host RAM)

273.4 GB required256.0 GB available

107% VRAM needed

17.4 GB over capacity — needs offload or smaller quantization

Fit status

Runs with offload (needs ~15.6 GB host RAM)

Decode

37.9 tok/s

TTFT

5109 ms

Safe context

Memory

273.4 GB / 256.0 GB

Offload

10%

Memory breakdown

Weights244.0 GB

KV Cache2.9 GB

Runtime0.9 GB

Headroom25.6 GB

See how fast it feels

See how fast it feelsLlama 4 Maverick 17B 128E on AMD Instinct MI325X 256GB
1st promptCold start — includes initialization
>Solve: If a train travels at 60 mph for 2.5 hours, how far does it go?
Let me work through this step by step.

Given information:
  - Speed = 60 mph (miles per hour)
  - Time  = 2.5 hours

Formula: Distance = Speed x Time

Calculation:
  Distance = 60 mph x 2.5 h
           = 150 miles

The train travels 150 miles.

Double-check: at 60 mph the train covers 60 miles each hour. In half an hour it covers 30 miles. So in 2.5 hours: 60 + 60 + 30 = 150 miles. Confirmed.
2nd promptWarm — model ready, no init overhead
>What if the train stops for 20 minutes halfway?
The stop doesn't change the total distance covered while moving, but it affects total trip time.

Breakdown:
  - First half: 1.25 hours at 60 mph = 75 miles
  - Stop: 20 minutes (no distance)
  - Second half: 1.25 hours at 60 mph = 75 miles

Total distance: 150 miles (same)
Total time: 2.5 h + 0.33 h = 2.83 hours
Average speed: 150 / 2.83 = 53 mph

The distance stays the same but average speed drops to 53 mph because of the stop.
Estimated: 37.9 tok/s decode · 5.1s TTFT (warm) · 95 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 15.6 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	A	Runs with offload (needs ~14.3 GB host RAM)	38.3 tok/s	2755 ms	4K
Coding	A	Runs with offload (needs ~15.6 GB host RAM)	37.9 tok/s	5109 ms	4K
Agentic Coding	A	Runs with offload (needs ~18 GB host RAM)	37.1 tok/s	7599 ms	4K
Reasoning	A	Runs with offload (needs ~15.6 GB host RAM)	37.9 tok/s	6037 ms	4K
RAG	A	Runs with offload (needs ~18 GB host RAM)	37.1 tok/s	9499 ms	4K

Inference speed

Llama 4 Maverick 17B 128E inference speed — tokens per second by GPU & Mac

Estimated decode speed (tokens/sec) for Llama 4 Maverick 17B 128E at Q4_K_M across popular GPUs and Apple Silicon, including multi-GPU rigs, using the fastest local runtime per device. Fastest is Mac Studio M3 Ultra 256GB at ~7 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?
Mac Studio M3 Ultra 256GB	256 GB	Q4_K_M	6.7	Too big
Mac Studio M2 Ultra 128GB	128 GB	Q4_K_M	4.3	Too big
Mac Studio M1 Ultra 128GB	128 GB	Q4_K_M	4.1	Too big
MacBook Pro M4 Max 128GB	128 GB	Q4_K_M	3.2	Too big
MacBook Pro M4 Max 64GB	64 GB	Q4_K_M	3.2	Too big
RTX 5090 32GB	32 GB	Q4_K_M	2.4	Too big
2× RX 7900 XTX 24GB	48 GB	Q4_K_M	2.3	Too big
MacBook Pro M3 Max 64GB	64 GB	Q4_K_M	2.2	Too big
RTX 4090 24GB	24 GB	Q4_K_M	2.0	Too big
RTX 4080 Super 16GB	16 GB	Q4_K_M	2.0	Too big
RTX 3090 24GB	24 GB	Q4_K_M	2.0	Too big
RTX 4070 12GB	12 GB	Q4_K_M	2.0	Too big
RTX 3060 12GB	12 GB	Q4_K_M	2.0	Too big
RTX 4060 8GB	8 GB	Q4_K_M	2.0	Too big
RX 7900 XTX 24GB	24 GB	Q4_K_M	2.0	Too big
MacBook Pro M1 Max 64GB	64 GB	Q4_K_M	2.0	Too big
MacBook Pro M4 Pro 48GB	48 GB	Q4_K_M	2.0	Too big
2× RTX 4090 24GB	48 GB	Q4_K_M	2.0	Too big
2× RTX 3090 24GB	48 GB	Q4_K_M	2.0	Too big
4× RTX 3060 12GB	48 GB	Q4_K_M	2.0	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 Llama 4 Maverick 17B 128E (400B params) fits at each quantization level on AMD Instinct MI325X 256GB (256.0 GB usable).

Quant	Bits	VRAM	Quality	Fit
Q2_K	2	156.0 GB	Low	A82
Q3_K_SBest for your GPU	3	196.0 GB	Low	A82
NVFP4	4	224.0 GB	Medium	F0
Q4_K_M	4	244.0 GB	Medium	F0
Q5_K_M	5	288.0 GB	High	F0
Q6_K	6	328.0 GB	High	F0
Q8_0	8	428.0 GB	Very High	F0
F16	16	820.0 GB	Maximum	F0

Get started

Copy-paste commands to run Llama 4 Maverick 17B 128E on your machine.

Run

lms load Llama-4-Maverick-17B-128E-Instruct && lms server start

Frequently asked questions

Can AMD Instinct MI325X 256GB run Llama 4 Maverick 17B 128E?

Yes, AMD Instinct MI325X 256GB can run Llama 4 Maverick 17B 128E with a A grade (Runs with offload (needs ~15.6 GB host RAM)). Expected decode speed: 37.9 tok/s.

How much VRAM does Llama 4 Maverick 17B 128E need?

Llama 4 Maverick 17B 128E (400B parameters) requires approximately 273.4 GB of memory with Q4_K_M quantization.

What is the best quantization for Llama 4 Maverick 17B 128E?

The recommended quantization for Llama 4 Maverick 17B 128E is Q4_K_M, which balances quality and memory efficiency.

What speed will Llama 4 Maverick 17B 128E run at on AMD Instinct MI325X 256GB?

On AMD Instinct MI325X 256GB, Llama 4 Maverick 17B 128E achieves approximately 37.9 tokens per second decode speed with a time-to-first-token of 5109ms using Q4_K_M quantization.

Can AMD Instinct MI325X 256GB run Llama 4 Maverick 17B 128E for coding?

For coding workloads, Llama 4 Maverick 17B 128E on AMD Instinct MI325X 256GB receives a A grade with 37.9 tok/s and 4K context.

What context window can Llama 4 Maverick 17B 128E use on AMD Instinct MI325X 256GB?

On AMD Instinct MI325X 256GB, Llama 4 Maverick 17B 128E can safely use up to 4K tokens of context. The model's official context limit is 1.0M, but available memory constrains the safe maximum.

What should I upgrade first if Llama 4 Maverick 17B 128E 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.

See all results for AMD Instinct MI325X 256GB See all hardware for Llama 4 Maverick 17B 128E

Embed this result▼

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

<iframe src="https://willitrunai.com/embed/llama-4-maverick-17b-128e-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: