Can Devstral Small 2 24B Instruct run on Gaudi 3 128GB?

YES — Runs Great

S88Excellent

Estimated from fit model

Devstral Small 2 24B Instruct needs ~30.8 GB VRAM. Gaudi 3 128GB has 128.0 GB. With Q4_K_M quantization, expect ~190 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) — 30.8 GB, 190.2 tok/s, Runs well

30.8 GB required128.0 GB available

24% VRAM used

Fit status

Runs well

Decode

190.2 tok/s

TTFT

1018 ms

Safe context

256K

Memory

30.8 GB / 128.0 GB

Memory breakdown

Weights14.6 GB

KV Cache2.4 GB

Runtime0.9 GB

Headroom12.8 GB

See how fast it feels

See how fast it feelsDevstral Small 2 24B Instruct on Gaudi 3 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: 190.2 tok/s decode · 1.0s TTFT (warm) · 476 tok/s prefill

What limits this setup

The raw memory story may look fine, but the software ecosystem is still a constraint here.

Runtime ecosystem is narrower than CUDA

Intel GPUs can look attractive on memory per dollar, but local AI tooling, kernels, and model coverage are still broader and easier on CUDA today.

Best improvement path

Prefer CUDA if you want the path of least resistance

If your goal is maximum runtime coverage, easier troubleshooting, and better support for new local AI releases, CUDA is usually still the safer upgrade path.

Performance by workload

Workload	Grade	Fit	Decode	TTFT	Context
Chat	S	Runs well	190.2 tok/s	555 ms	256K
Coding	S	Runs well	190.2 tok/s	1018 ms	256K
Agentic Coding	S	Runs well	190.2 tok/s	1481 ms	256K
Reasoning	S	Runs well	190.2 tok/s	1203 ms	256K
RAG	S	Runs well	190.2 tok/s	1851 ms	256K

Inference speed

Devstral Small 2 24B Instruct inference speed — tokens per second by GPU & Mac

Estimated decode speed (tokens/sec) for Devstral Small 2 24B Instruct at Q4_K_M across popular GPUs and Apple Silicon, using the fastest local runtime per device. Fastest is RTX 5090 32GB at ~88 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	88.2	Fits
RTX 4090 24GB	24 GB	Q4_K_M	56.3	Tight
RX 7900 XTX 24GB	24 GB	Q4_K_M	50.8	Tight
RTX 3090 24GB	24 GB	Q4_K_M	48.1	Tight
Mac Studio M3 Ultra 256GB	256 GB	Q4_K_M	40.9	Fits
MacBook Pro M4 Max 128GB	128 GB	Q4_K_M	36.8	Fits
MacBook Pro M4 Max 64GB	64 GB	Q4_K_M	36.8	Fits
Mac Studio M2 Ultra 128GB	128 GB	Q4_K_M	34.1	Fits
Mac Studio M1 Ultra 128GB	128 GB	Q4_K_M	32.3	Fits
MacBook Pro M4 Pro 48GB	48 GB	Q4_K_M	23.2	Fits
RTX 4080 Super 16GB	16 GB	Q4_K_M	21.3	Too big
MacBook Pro M3 Max 64GB	64 GB	Q4_K_M	17.6	Fits
MacBook Pro M1 Max 64GB	64 GB	Q4_K_M	16.2	Fits
RTX 4070 12GB	12 GB	Q4_K_M	7.5	Too big
RTX 3060 12GB	12 GB	Q4_K_M	4.7	Too big
RTX 4060 8GB	8 GB	Q4_K_M	2.2	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 Devstral Small 2 24B Instruct (24B params) fits at each quantization level on Gaudi 3 128GB (128.0 GB usable).

Quant	Bits	VRAM	Quality	Fit
Q2_K	2	9.4 GB	Low	A80
Q3_K_S	3	11.8 GB	Low	A80
NVFP4	4	13.4 GB	Medium	A80
Q4_K_M	4	14.6 GB	Medium	A80
Q5_K_M	5	17.3 GB	High	A80
Q6_K	6	19.7 GB	High	A80
Q8_0	8	25.7 GB	Very High	A81
F16Best for your GPU	16	49.2 GB	Maximum	A85

Get started

Copy-paste commands to run Devstral Small 2 24B Instruct on your machine.

Run

ollama run devstral-small-2

Your hardware

More models your Gaudi 3 128GB can run

Model	Params	Grade	Decode
Devstral 2 123B Instruct	123B	S	37.5 tok/s
Qwen3-Coder 30B A3B Instruct	30.5B	S	391.6 tok/s
Qwen 3.5 27B	27B	S	169.8 tok/s
Qwen 3.6 27B	27B	S	105.9 tok/s
Qwen 3.5 122B A10B	122B	S	104.1 tok/s

Frequently asked questions

Can Gaudi 3 128GB run Devstral Small 2 24B Instruct?

Yes, Gaudi 3 128GB can run Devstral Small 2 24B Instruct with a S grade (Runs well). Expected decode speed: 190.2 tok/s.

How much VRAM does Devstral Small 2 24B Instruct need?

Devstral Small 2 24B Instruct (24B parameters) requires approximately 30.8 GB of memory with Q4_K_M quantization.

What is the best quantization for Devstral Small 2 24B Instruct?

The recommended quantization for Devstral Small 2 24B Instruct is Q4_K_M, which balances quality and memory efficiency.

What speed will Devstral Small 2 24B Instruct run at on Gaudi 3 128GB?

On Gaudi 3 128GB, Devstral Small 2 24B Instruct achieves approximately 190.2 tokens per second decode speed with a time-to-first-token of 1018ms using Q4_K_M quantization.

Can Gaudi 3 128GB run Devstral Small 2 24B Instruct for coding?

For coding workloads, Devstral Small 2 24B Instruct on Gaudi 3 128GB receives a S grade with 190.2 tok/s and 256K context.

What context window can Devstral Small 2 24B Instruct use on Gaudi 3 128GB?

On Gaudi 3 128GB, Devstral Small 2 24B 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.

What should I upgrade first if Devstral Small 2 24B Instruct feels slow on Gaudi 3 128GB?

Prefer CUDA if you want the path of least resistance. If your goal is maximum runtime coverage, easier troubleshooting, and better support for new local AI releases, CUDA is usually still the safer upgrade path.

Would CUDA be a better path than Gaudi 3 128GB for Devstral Small 2 24B Instruct?

Often yes, if your goal is the easiest setup and the widest runtime support. Intel can offer attractive memory capacity, but CUDA still tends to win on tooling maturity, guides, kernels, and model coverage for local AI.

See all results for Gaudi 3 128GB See all hardware for Devstral Small 2 24B Instruct

Embed this result▼

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

<iframe src="https://willitrunai.com/embed/devstral-small-2-24b-on-gaudi-3-128gb" 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

190.2 tok/s

555 ms

256K

Coding

Runs well

190.2 tok/s

1018 ms

256K

Agentic Coding

Runs well

190.2 tok/s

1481 ms

256K

Reasoning

Runs well

190.2 tok/s

1203 ms

256K

RAG

Runs well

190.2 tok/s

1851 ms

256K

Devstral Small 2 24B Instruct inference speed — tokens per second by GPU & Mac

GPU / Mac

Memory

Quant

Speed (tok/s)

Fits?

RTX 5090 32GB

32 GB

Q4_K_M

88.2

Fits

RTX 4090 24GB

24 GB

Q4_K_M

56.3

Tight

RX 7900 XTX 24GB

24 GB

Q4_K_M

50.8

Tight

RTX 3090 24GB

24 GB

Q4_K_M

48.1

Tight

Mac Studio M3 Ultra 256GB

256 GB

Q4_K_M

40.9

Fits

MacBook Pro M4 Max 128GB

128 GB

Q4_K_M

36.8

Fits

MacBook Pro M4 Max 64GB

64 GB

Q4_K_M

36.8

Fits

Mac Studio M2 Ultra 128GB

128 GB

Q4_K_M

34.1

Fits

Mac Studio M1 Ultra 128GB

128 GB

Q4_K_M

32.3

Fits

MacBook Pro M4 Pro 48GB

48 GB

Q4_K_M

23.2

Fits

RTX 4080 Super 16GB

16 GB

Q4_K_M

21.3

Too big

MacBook Pro M3 Max 64GB

64 GB

Q4_K_M

17.6

Fits

MacBook Pro M1 Max 64GB

64 GB

Q4_K_M

16.2

Fits

RTX 4070 12GB

12 GB

Q4_K_M

7.5

Too big

RTX 3060 12GB

12 GB

Q4_K_M

4.7

Too big

RTX 4060 8GB

8 GB

Q4_K_M

2.2

Too big

Quantization options

How Devstral Small 2 24B Instruct (24B params) fits at each quantization level on Gaudi 3 128GB (128.0 GB usable).

Quant	Bits	VRAM	Quality	Fit
Q2_K	2	9.4 GB	Low	A80
Q3_K_S	3	11.8 GB	Low	A80
NVFP4	4	13.4 GB	Medium	A80
Q4_K_M	4	14.6 GB	Medium	A80
Q5_K_M	5	17.3 GB	High	A80
Q6_K	6	19.7 GB	High	A80
Q8_0	8	25.7 GB	Very High	A81
F16Best for your GPU	16	49.2 GB	Maximum	A85

Model

Params

Grade

Decode

Capabilities

Devstral 2 123B Instruct

123B

37.5 tok/s

Qwen3-Coder 30B A3B Instruct

30.5B

391.6 tok/s

Qwen 3.5 27B

27B

169.8 tok/s

Qwen 3.6 27B

27B

105.9 tok/s

Qwen 3.5 122B A10B

122B

104.1 tok/s