Is unified memory on Mac mini M2 24GB as fast as VRAM for Devstral Small 1.1?

Not always. Mac mini M2 24GB can often fit larger models thanks to unified memory, but a discrete GPU with dedicated high-bandwidth VRAM may still decode faster once the model fits. For this combination, the important distinction is capacity versus sustained throughput.

Can Devstral Small 1.1 run on Mac mini M2 24GB?

Q: What should I upgrade first if Devstral Small 1.1 feels slow on Mac mini M2 24GB?

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.

BARELY — Tight on Memory

B67Good

Estimated from fit model

Devstral Small 1.1 needs ~20.6 GB VRAM. Mac mini M2 24GB has 17.3 GB. With Q4_K_M quantization, expect ~3 tok/s.

Runtime: llama.cppCapacity: OffloadBandwidth: Very lowStack: 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) — 20.6 GB, 3.7 tok/s, Very compromised (needs ~2.3 GB host RAM)

20.6 GB required17.3 GB available

119% VRAM needed

3.3 GB over capacity — needs offload or smaller quantization

Fit status

Very compromised (needs ~2.3 GB host RAM)

Decode

3.7 tok/s

TTFT

52742 ms

Safe context

Memory

20.6 GB / 17.3 GB

Offload

20%

Memory breakdown

Weights14.6 GB

KV Cache2.4 GB

Runtime0.9 GB

Headroom2.6 GB

See how fast it feels

See how fast it feelsDevstral Small 1.1 on Mac mini M2 24GB
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: 3.7 tok/s decode · 52.7s TTFT (warm) · 9 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 20% 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.

Shared-memory contention still exists

The OS, browser, and inference runtime all compete for the same physical memory pool, so real-world headroom is less forgiving than raw capacity suggests.

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 {ram} 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	3.7 tok/s	28567 ms	4K
Coding	B	Very compromised	3.4 tok/s	56698 ms	4K
Agentic Coding	F	Too heavy	3.0 tok/s	94525 ms	4K
Reasoning	B	Very compromised	3.4 tok/s	67006 ms	4K
RAG	F	Too heavy	3.0 tok/s	118156 ms	4K

Quantization options

How Devstral Small 1.1 (24B params) fits at each quantization level on Mac mini M2 24GB (17.3 GB usable).

Quant	Bits	VRAM	Quality	Fit
Q2_K	2	9.4 GB	Low	S91
Q3_K_SBest for your GPU	3	11.8 GB	Low	S90

Get started

Copy-paste commands to run Devstral Small 1.1 on your machine.

Run

lms load Devstral-Small-2507 && lms server start

Upgrade options

Hardware that runs Devstral Small 1.1 well

MacBook Pro M4 32GBBudget pick

32 GB Unified (+8)120 GB/s (+20)

Removes host-memory offload, which is usually the single biggest latency and throughput win.9.5 tok/s decode

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

Raises estimated decode speed by about 157%.

~$799 MSRP

Mac mini M4 32GBBest value

32 GB Unified (+8)120 GB/s (+20)

Removes host-memory offload, which is usually the single biggest latency and throughput win.9.5 tok/s decode

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

Raises estimated decode speed by about 157%.

~$1,099 MSRP

Mac mini M4 64GBApple upgrade

64 GB Unified (+40)120 GB/s (+20)

Removes host-memory offload, which is usually the single biggest latency and throughput win.9.5 tok/s decode

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

Raises estimated decode speed by about 157%.

~$1,099 MSRP

RTX 5090 32GBBiggest leap

32 GB VRAM (+8)1792 GB/s (+1692)

Removes host-memory offload, which is usually the single biggest latency and throughput win.62 tok/s decode

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

Raises estimated decode speed by about 1576%.

~$1,999 MSRP

Frequently asked questions

Can Mac mini M2 24GB run Devstral Small 1.1?

Yes, Mac mini M2 24GB can run Devstral Small 1.1 with a B grade (Very compromised). Expected decode speed: 3.4 tok/s.

How much VRAM does Devstral Small 1.1 need?

Devstral Small 1.1 (24B parameters) requires approximately 20.6 GB of memory with Q4_K_M quantization.

What is the best quantization for Devstral Small 1.1?

The recommended quantization for Devstral Small 1.1 is Q4_K_M, which balances quality and memory efficiency.

What speed will Devstral Small 1.1 run at on Mac mini M2 24GB?

On Mac mini M2 24GB, Devstral Small 1.1 achieves approximately 3.4 tokens per second decode speed with a time-to-first-token of 56698ms using Q4_K_M quantization.

Can Mac mini M2 24GB run Devstral Small 1.1 for coding?

For coding workloads, Devstral Small 1.1 on Mac mini M2 24GB receives a B grade with 3.4 tok/s and 4K context.

What context window can Devstral Small 1.1 use on Mac mini M2 24GB?

On Mac mini M2 24GB, Devstral Small 1.1 can safely use up to 4K tokens of context. The model's official context limit is 131K, but available memory constrains the safe maximum.

What should I upgrade first if Devstral Small 1.1 feels slow on Mac mini M2 24GB?

See all results for Mac mini M2 24GB See all hardware for Devstral Small 1.1

Embed this result▼

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

<iframe src="https://willitrunai.com/embed/devstral-small-2507-on-m2-24gb" width="400" height="180" frameborder="0" style="border:none;border-radius:12px;overflow:hidden;" title="Will It Run AI — fit result"></iframe>

Preview: