Can Devstral Small 2 24B Instruct run on RX 6900 XT 16GB?

Q: What should I upgrade first if Devstral Small 2 24B Instruct feels slow on RX 6900 XT 16GB?

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 NVFP4

A79Great

Estimated from fit model

Devstral Small 2 24B Instruct needs ~18.4 GB VRAM. RX 6900 XT 16GB has 16.0 GB. With NVFP4 quantization, expect ~14 tok/s.

Runtime: llama.cppCapacity: OffloadBandwidth: MediumStack: 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.

Devstral Small 2 24B Instruct at Q4_K_M needs 19.6 GB — too much for RX 6900 XT 16GB (16.0 GB). Runs at NVFP4 (18.4 GB) with medium quality. 3 quantization levels fit.

Capabilities:

Select quantization to explore

Q4_K_M (Medium quality) — 19.6 GB, exceeds 16.0 GB available

19.6 GB required16.0 GB available

123% VRAM needed

3.6 GB over capacity — needs offload or smaller quantization

Fit status

Too heavy

Decode

10.5 tok/s

TTFT

18429 ms

Safe context

Memory

19.6 GB / 16.0 GB

Offload

20%

Memory breakdown

Weights14.6 GB

KV Cache2.4 GB

Runtime0.9 GB

Headroom1.6 GB

See how fast it feels

With memory offload — actual speed may be lower

See how fast it feelsDevstral Small 2 24B Instruct on RX 6900 XT 16GB
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: 10.5 tok/s decode · 18.4s TTFT (warm) · 26 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 1.7 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	Very compromised (needs ~1.9 GB host RAM)	12.0 tok/s	8778 ms	4K
Coding	F	Too heavy	9.8 tok/s	19811 ms	4K
Agentic Coding	F	Too heavy	8.2 tok/s	34327 ms	4K
Reasoning	F	Too heavy	10.5 tok/s	21779 ms	4K
RAG	F	Too heavy	8.2 tok/s	42909 ms	4K

Quantization options

How Devstral Small 2 24B Instruct (24B params) fits at each quantization level on RX 6900 XT 16GB (16.0 GB usable).

Quant	Bits	VRAM	Quality	Fit
Q2_K	2	9.4 GB	Low	S92
Q3_K_SBest for your GPU	3	11.8 GB	Low	S92

Get started

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

Run

ollama run devstral-small-2

Upgrade options

Hardware that runs Devstral Small 2 24B Instruct well

RX 7900 XT 20GBBudget pick

20 GB VRAM (+4)800 GB/s (+288)

Makes the model fit on the accelerator instead of staying completely out of reach.35.2 tok/s decode

Makes the model fit on the accelerator instead of staying completely out of reach.

Removes host-memory offload, which is usually the single biggest latency and throughput win.

~$899 MSRP

RX 7900 XTX 24GBBest value

24 GB VRAM (+8)960 GB/s (+448)

Makes the model fit on the accelerator instead of staying completely out of reach.50.8 tok/s decode

Makes the model fit on the accelerator instead of staying completely out of reach.

Removes host-memory offload, which is usually the single biggest latency and throughput win.

~$999 MSRP

Radeon AI PRO R9700 32GBAMD upgrade

32 GB VRAM (+16)640 GB/s (+128)

Makes the model fit on the accelerator instead of staying completely out of reach.27.7 tok/s decode

Makes the model fit on the accelerator instead of staying completely out of reach.

Removes host-memory offload, which is usually the single biggest latency and throughput win.

~$1,899 MSRP

Frequently asked questions

Can RX 6900 XT 16GB run Devstral Small 2 24B Instruct?

Yes, RX 6900 XT 16GB can run Devstral Small 2 24B Instruct at NVFP4 quantization (Very compromised (needs ~1.7 GB host RAM)). The recommended Q4_K_M requires 19.6 GB which exceeds available memory, but at NVFP4 it needs only 18.4 GB. Expected decode speed: 13.7 tok/s.

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

Devstral Small 2 24B Instruct (24B parameters) requires approximately 19.6 GB at Q4_K_M quantization. On RX 6900 XT 16GB, it fits at NVFP4 using 18.4 GB.

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

The recommended quantization is Q4_K_M, but on RX 6900 XT 16GB the best fitting quantization is NVFP4, which uses 18.4 GB.

What speed will Devstral Small 2 24B Instruct run at on RX 6900 XT 16GB?

On RX 6900 XT 16GB, Devstral Small 2 24B Instruct achieves approximately 13.7 tokens per second decode speed with a time-to-first-token of 14104ms using NVFP4 quantization.

Can RX 6900 XT 16GB run Devstral Small 2 24B Instruct for coding?

For coding workloads, Devstral Small 2 24B Instruct on RX 6900 XT 16GB receives a F grade with 9.8 tok/s and 4K context.

What context window can Devstral Small 2 24B Instruct use on RX 6900 XT 16GB?

On RX 6900 XT 16GB, Devstral Small 2 24B Instruct can safely use up to 4K tokens of context at NVFP4 quantization. The model's official context limit is 256K, but available memory constrains the safe maximum.

See all results for RX 6900 XT 16GB 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-rx-6900-xt-16gb" width="400" height="180" frameborder="0" style="border:none;border-radius:12px;overflow:hidden;" title="Will It Run AI — fit result"></iframe>

Preview: