Can Laguna S 2.1 run on AMD Instinct MI250X 128GB?

YES — Runs Great

S92Excellent

Estimated from fit model

Laguna S 2.1 needs ~86.2 GB VRAM. AMD Instinct MI250X 128GB has 128.0 GB. With Q4_K_M quantization, expect ~106 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) — 86.2 GB, 106.4 tok/s, Runs well

86.2 GB required128.0 GB available

67% VRAM used

Fit status

Runs well

Decode

106.4 tok/s

TTFT

1820 ms

Safe context

930K

Memory

86.2 GB / 128.0 GB

Memory breakdown

Weights71.7 GB

KV Cache0.7 GB

Runtime0.9 GB

Headroom12.8 GB

See how fast it feels

See how fast it feelsLaguna S 2.1 on AMD Instinct MI250X 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: 106.4 tok/s decode · 1.8s TTFT (warm) · 266 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	106.4 tok/s	993 ms	930K
Coding	S	Runs well	106.4 tok/s	1820 ms	930K
Agentic Coding	S	Runs well	106.4 tok/s	2648 ms	930K
Reasoning	S	Runs well	106.4 tok/s	2151 ms	930K
RAG	S	Runs well	106.4 tok/s	3310 ms	930K

Inference speed

Laguna S 2.1 inference speed — tokens per second by GPU & Mac

Estimated decode speed (tokens/sec) for Laguna S 2.1 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 ~37 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	36.8	Fits
Mac Studio M2 Ultra 128GB	128 GB	Q4_K_M	30.6	Tight
Mac Studio M1 Ultra 128GB	128 GB	Q4_K_M	29.1	Tight
MacBook Pro M4 Max 128GB	128 GB	Q4_K_M	22.7	Tight
2× RX 7900 XTX 24GB	48 GB	Q4_K_M	13.5	Too big
MacBook Pro M4 Max 64GB	64 GB	Q4_K_M	11.2	Too big
2× RTX 4090 24GB	48 GB	Q4_K_M	9.0	Too big
MacBook Pro M3 Max 64GB	64 GB	Q4_K_M	7.8	Too big
RTX 5090 32GB	32 GB	Q4_K_M	7.7	Too big
2× RTX 3090 24GB	48 GB	Q4_K_M	7.7	Too big
MacBook Pro M1 Max 64GB	64 GB	Q4_K_M	7.2	Too big
4× RTX 3060 12GB	48 GB	Q4_K_M	6.8	Too big
MacBook Pro M4 Pro 48GB	48 GB	Q4_K_M	6.2	Too big
RTX 4090 24GB	24 GB	Q4_K_M	4.9	Too big
RX 7900 XTX 24GB	24 GB	Q4_K_M	4.4	Too big
RTX 3090 24GB	24 GB	Q4_K_M	4.2	Too big
RTX 4080 Super 16GB	16 GB	Q4_K_M	3.9	Too big
RTX 4070 12GB	12 GB	Q4_K_M	2.4	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

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 Laguna S 2.1 (117.5999984741211B params) fits at each quantization level on AMD Instinct MI250X 128GB (128.0 GB usable).

Quant	Bits	VRAM	Quality	Fit
Q1_0_G128	1.125	16.9 GB	Very Low	A75
Q2_0_G128	1.71	31.4 GB	Low	A77
Q2_K	2	45.9 GB	Low	A80
Q3_K_S	3	57.6 GB	Low	A81
NVFP4	4	65.9 GB	Medium	A83
Q4_K_M	4	71.7 GB	Medium	A83
Q5_K_M	5	84.7 GB	High	A83
Q6_KBest for your GPU	6	96.4 GB	High	A83
Q8_0	8	125.8 GB	Very High	F0
F16	16	241.1 GB	Maximum	F0

Get started

Copy-paste commands to run Laguna S 2.1 on your machine.

Run

docker run --rm -it ghcr.io/ggerganov/llama.cpp:full \
  --hf-repo "poolside/Laguna-S-2.1" \
  --hf-file "Laguna-S-2.1-Q4_K_M.gguf" \
  -c 4096 -ngl 99

Your hardware

More models your AMD Instinct MI250X 128GB can run

Model	Params	Grade	Decode
Devstral 2 123B Instruct	123B	S	36.2 tok/s
Qwen 3.5 122B A10B	122B	S	100.3 tok/s
Pixtral Large 124B	124B	S	35.9 tok/s

Frequently asked questions

Can AMD Instinct MI250X 128GB run Laguna S 2.1?

Yes, AMD Instinct MI250X 128GB can run Laguna S 2.1 with a S grade (Runs well). Expected decode speed: 106.4 tok/s.

How much VRAM does Laguna S 2.1 need?

Laguna S 2.1 (117.5999984741211B parameters) requires approximately 86.2 GB of memory with Q4_K_M quantization.

What is the best quantization for Laguna S 2.1?

The recommended quantization for Laguna S 2.1 is Q4_K_M, which balances quality and memory efficiency.

What speed will Laguna S 2.1 run at on AMD Instinct MI250X 128GB?

On AMD Instinct MI250X 128GB, Laguna S 2.1 achieves approximately 106.4 tokens per second decode speed with a time-to-first-token of 1820ms using Q4_K_M quantization.

Can AMD Instinct MI250X 128GB run Laguna S 2.1 for coding?

For coding workloads, Laguna S 2.1 on AMD Instinct MI250X 128GB receives a S grade with 106.4 tok/s and 930K context.

What context window can Laguna S 2.1 use on AMD Instinct MI250X 128GB?

On AMD Instinct MI250X 128GB, Laguna S 2.1 can safely use up to 930K tokens of context. The model's official context limit is 1.0M, but available memory constrains the safe maximum.

See all results for AMD Instinct MI250X 128GB See all hardware for Laguna S 2.1

Embed this result▼

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

<iframe src="https://willitrunai.com/embed/laguna-s-2.1-on-instinct-mi250x-128gb" width="400" height="180" frameborder="0" style="border:none;border-radius:12px;overflow:hidden;" title="Will It Run AI — fit result"></iframe>

Preview:

Workload

Grade

Fit

Decode

TTFT

Context

Chat

Runs well

106.4 tok/s

993 ms

930K

Coding

Runs well

106.4 tok/s

1820 ms

930K

Agentic Coding

Runs well

106.4 tok/s

2648 ms

930K

Reasoning

Runs well

106.4 tok/s

2151 ms

930K

RAG

Runs well

106.4 tok/s

3310 ms

930K

Laguna S 2.1 inference speed — tokens per second by GPU & Mac

GPU / Mac

Memory

Quant

Speed (tok/s)

Fits?

Mac Studio M3 Ultra 256GB

256 GB

Q4_K_M

36.8

Fits

Mac Studio M2 Ultra 128GB

128 GB

Q4_K_M

30.6

Tight

Mac Studio M1 Ultra 128GB

128 GB

Q4_K_M

29.1

Tight

MacBook Pro M4 Max 128GB

128 GB

Q4_K_M

22.7

Tight

2× RX 7900 XTX 24GB

48 GB

Q4_K_M

13.5

Too big

MacBook Pro M4 Max 64GB

64 GB

Q4_K_M

11.2

Too big

2× RTX 4090 24GB

48 GB

Q4_K_M

9.0

Too big

MacBook Pro M3 Max 64GB

64 GB

Q4_K_M

7.8

Too big

RTX 5090 32GB

32 GB

Q4_K_M

7.7

Too big

2× RTX 3090 24GB

48 GB

Q4_K_M

7.7

Too big

MacBook Pro M1 Max 64GB

64 GB

Q4_K_M

7.2

Too big

4× RTX 3060 12GB

48 GB

Q4_K_M

6.8

Too big

MacBook Pro M4 Pro 48GB

48 GB

Q4_K_M

6.2

Too big

RTX 4090 24GB

24 GB

Q4_K_M

4.9

Too big

RX 7900 XTX 24GB

24 GB

Q4_K_M

4.4

Too big

RTX 3090 24GB

24 GB

Q4_K_M

4.2

Too big

RTX 4080 Super 16GB

16 GB

Q4_K_M

3.9

Too big

RTX 4070 12GB

12 GB

Q4_K_M

2.4

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

Quantization options

How Laguna S 2.1 (117.5999984741211B params) fits at each quantization level on AMD Instinct MI250X 128GB (128.0 GB usable).

Quant	Bits	VRAM	Quality	Fit
Q1_0_G128	1.125	16.9 GB	Very Low	A75
Q2_0_G128	1.71	31.4 GB	Low	A77
Q2_K	2	45.9 GB	Low	A80
Q3_K_S	3	57.6 GB	Low	A81
NVFP4	4	65.9 GB	Medium	A83
Q4_K_M	4	71.7 GB	Medium	A83
Q5_K_M	5	84.7 GB	High	A83
Q6_KBest for your GPU	6	96.4 GB	High	A83
Q8_0	8	125.8 GB	Very High	F0
F16	16	241.1 GB	Maximum	F0

Model

Params

Grade

Decode

Capabilities

Devstral 2 123B Instruct

123B

36.2 tok/s

Qwen 3.5 122B A10B

122B

100.3 tok/s

Pixtral Large 124B

124B

35.9 tok/s

Frequently asked questions

Can AMD Instinct MI250X 128GB run Laguna S 2.1?

Yes, AMD Instinct MI250X 128GB can run Laguna S 2.1 with a S grade (Runs well). Expected decode speed: 106.4 tok/s.

How much VRAM does Laguna S 2.1 need?

Laguna S 2.1 (117.5999984741211B parameters) requires approximately 86.2 GB of memory with Q4_K_M quantization.

What is the best quantization for Laguna S 2.1?

The recommended quantization for Laguna S 2.1 is Q4_K_M, which balances quality and memory efficiency.

What speed will Laguna S 2.1 run at on AMD Instinct MI250X 128GB?

On AMD Instinct MI250X 128GB, Laguna S 2.1 achieves approximately 106.4 tokens per second decode speed with a time-to-first-token of 1820ms using Q4_K_M quantization.

Can AMD Instinct MI250X 128GB run Laguna S 2.1 for coding?

For coding workloads, Laguna S 2.1 on AMD Instinct MI250X 128GB receives a S grade with 106.4 tok/s and 930K context.

What context window can Laguna S 2.1 use on AMD Instinct MI250X 128GB?

On AMD Instinct MI250X 128GB, Laguna S 2.1 can safely use up to 930K tokens of context. The model's official context limit is 1.0M, but available memory constrains the safe maximum.