Outmake Models
Models

Devstral Small 2 24B

Mistral AI · November 2025

GetHugging Face

For powering coding agents that use tools to explore codebases and edit multiple files.

Rating, estimated
3.4
Downloads
162K
Parameters
24B
Context
256K
Download
6.4 to 48 GB
Commercial use
Yes

Runs well on M4 · 16 GB

Outmake picks the Q2_K_L size: a 9.9 GB download that uses 11 GB of memory, and writes about 4.8 words (6.4 tokens) a second.

About

How it compares

Of the 9 chat models Outmake recommends for an 18 GB Mac, it is the slowest, by Outmake's estimate.

Of those 9 models, it was the 7th most downloaded on Hugging Face last month.

Good to know

It shares its design with the Ministral 3 models.

Sources: mistralai/Devstral-Small-2-24B-Instruct-2512 model card, Outmake's speed estimates, LM Arena and Hugging Face. Every line comes from one of these.

Good at

  • Coding questions in chat
  • Calling tools
  • Long inputs
  • Reading pictures

Licence

Apache 2.0. Commercial use: yes. The maker's page has the full terms.

Devstral Small 2 24B runs well from a Mac with an M1 and 16 GB of memory, at Q2_K_L, writing about 2.8 words (3.7 tokens) a second. 56 of 59 Mac configurations run it well.

Compare with

The nearest models in size, judged for the same Mac.

Sizes

SizeDownloadMemory at 8KMemory at 32KOpen
Q3_K_MSmaller12 GB13 GB15 GBOpen
UD-Q4_K_XLBalanced15 GB16 GB18 GBOpen
Q6_KBest20 GB21 GB23 GBOpen
All 26 sizes
SizeDownloadMemory at 8KMemory at 32KOpen
UD-IQ1_S6.4 GB7.1 GBestimated9.3 GBOpen
UD-IQ1_M6.9 GB7.6 GBestimated9.7 GBOpen
UD-IQ2_XXS7.6 GB8.3 GBestimated10 GBOpen
UD-IQ2_M9.1 GB9.8 GBestimated12 GBOpen
Q2_K9.8 GB10 GBestimated13 GBOpen
Q2_K_L9.9 GB11 GBestimated13 GBOpen
UD-Q2_K_XL10 GB11 GBestimated13 GBOpen
UD-IQ3_XXS10 GB11 GBestimated13 GBOpen
Q3_K_S11 GB12 GBestimated14 GBOpen
UD-Q3_K_XL13 GB13 GBestimated16 GBOpen
IQ4_XS14 GB14 GBestimated16 GBOpen
IQ4_NL14 GB15 GBestimated17 GBOpen
Q4_014 GB15 GBestimated17 GBOpen
Q4_K_S14 GB15 GBestimated17 GBOpen
Q4_K_M15 GB16 GBestimated18 GBOpen
Q4_116 GB16 GBestimated18 GBOpen
Q5_K_S17 GB18 GBestimated20 GBOpen
Q5_K_M18 GB18 GBestimated20 GBOpen
UD-Q5_K_XL18 GB18 GBestimated20 GBOpen
UD-Q6_K_XL22 GB22 GBestimated24 GBOpen
Q8_026 GB26 GBestimated29 GBOpen
UD-Q8_K_XL30 GB30 GBestimated32 GBOpen
BF1648 GB48 GBestimated51 GBOpen

Memory is what llama.cpp's own calculator reported for each size when the catalog was built, with the app's settings. Smaller, Balanced and Best are the three sizes the store names.

Run It Anywhere

Devstral Small 2 24B is an open model, so any app that runs GGUF files can run it. These take the UD-Q4_K_XL size from unsloth straight from Hugging Face.

  • LM StudioMac, Windows, LinuxOpen
  • JanMac, Windows, LinuxOpen
  • MstyMac, Windows, LinuxOpen
  • OllamaMac, Windows, Linuxollama run hf.co/unsloth/Devstral-Small-2-24B-Instruct-2512-GGUF:UD-Q4_K_XL
  • llama.cppMac, Windows, Linuxllama-server -hf unsloth/Devstral-Small-2-24B-Instruct-2512-GGUF:UD-Q4_K_XL
  • Docker Model RunnerMac, Windows, Linuxdocker model run hf.co/unsloth/Devstral-Small-2-24B-Instruct-2512-GGUF:UD-Q4_K_XL
On a PC Graphics Card
Video memoryLargest size that fitsDownloadMemory at 8K
6 GBNone worth runningneeds 9.8 GB
8 GBNone worth runningneeds 9.8 GB
10 GBUD-IQ2_M9.1 GB9.8 GB
11 GBUD-Q2_K_XL10 GB11 GB
12 GBUD-IQ3_XXS10 GB11 GB
16 GBQ4_K_S14 GB15 GB
20 GBUD-Q5_K_XL18 GB18 GB
24 GBUD-Q6_K_XL22 GB22 GB
32 GBUD-Q8_K_XL30 GB30 GB
48 GBUD-Q8_K_XL30 GB30 GB

The largest size, of those that keep three bits a weight or more, whose memory at an 8K chat is under 92% of the card's; the rest is the desktop's. Estimated from the files, the same way as for a Mac; speed on a PC is not estimated yet. A size that doesn't fit the card still runs, slower, with part of it in ordinary memory.

On every Mac
MacSizeFitMemoryWritesReads
M6 · 16 GBQ2_K_LRuns well11 GB6.8262
M6 · 24 GBQ3_K_MRuns well15 GB5.8262
M6 · 32 GBQ3_K_MRuns great15 GB5.8262
M5 · 16 GBQ2_K_LRuns well11 GB6.297
M5 · 24 GBQ3_K_MRuns well15 GB5.297
M5 · 32 GBQ3_K_MRuns great15 GB5.297
M5 Pro · 24 GBQ3_K_MRuns well15 GB10193
M5 Pro · 48 GBQ6_KRuns great23 GB7.4193
M5 Pro · 64 GBQ6_KRuns great23 GB7.4242
M5 Max · 36 GBUD-Q4_K_XLRuns great16 GB13387
M5 Max · 48 GBQ6_KRuns great23 GB15484
M5 Max · 64 GBQ6_KRuns great23 GB15484
M5 Max · 128 GBQ6_KRuns great32 GB15484
M5 Ultra · 96 GBQ6_KRuns great32 GB28774
M5 Ultra · 256 GBQ6_KRuns great32 GB28967
M5 Ultra · 512 GBQ6_KRuns great32 GB28967
M4 · 16 GBQ2_K_LRuns well11 GB4.833
M4 · 24 GBQ3_K_MRuns well15 GB4.133
M4 · 32 GBQ3_K_MRuns great15 GB4.133
M4 Pro · 24 GBQ3_K_MRuns well15 GB965
M4 Pro · 48 GBQ6_KRuns great23 GB6.365
M4 Pro · 64 GBQ6_KRuns great23 GB6.365
M4 Max · 36 GBUD-Q4_K_XLRuns great16 GB11130
M4 Max · 48 GBQ6_KRuns great23 GB12163
M4 Max · 64 GBQ6_KRuns great23 GB12163
M4 Max · 96 GBQ6_KRuns great32 GB9.1130
M4 Max · 128 GBQ6_KRuns great32 GB12163
M3 · 8 GBQ3_K_MToo big24 GB
M3 · 16 GBQ2_K_LRuns well11 GB4.127
M3 · 24 GBQ3_K_MRuns well15 GB3.527
M3 Pro · 18 GBQ3_K_SRuns well12 GB5.547
M3 Pro · 36 GBUD-Q4_K_XLRuns great16 GB4.447
M3 Max · 36 GBUD-Q4_K_XLRuns great16 GB8.5101
M3 Max · 48 GBQ6_KRuns great23 GB9135
M3 Max · 64 GBQ6_KRuns great23 GB9135
M3 Max · 96 GBQ6_KRuns great32 GB6.9101
M3 Max · 128 GBQ6_KRuns great32 GB9135
M3 Ultra · 96 GBQ6_KRuns great32 GB17202
M3 Ultra · 256 GBQ6_KRuns great32 GB17202
M3 Ultra · 512 GBQ6_KRuns great32 GB17269
M2 · 8 GBQ3_K_MToo big24 GB
M2 · 16 GBQ2_K_LRuns well11 GB4.126
M2 · 24 GBQ3_K_MRuns well15 GB3.526
M2 Pro · 16 GBQ2_K_LRuns well11 GB7.953
M2 Pro · 32 GBQ3_K_MRuns great15 GB6.853
M2 Max · 32 GBQ3_K_MRuns great15 GB1399
M2 Max · 64 GBQ6_KRuns great23 GB999
M2 Max · 96 GBQ6_KRuns great32 GB9125
M2 Ultra · 64 GBQ6_KRuns great23 GB17198
M2 Ultra · 128 GBQ6_KRuns great32 GB17198
M2 Ultra · 192 GBQ6_KRuns great32 GB17198
M1 · 8 GBQ3_K_MToo big24 GB
M1 · 16 GBQ2_K_LRuns well11 GB2.821
M1 Pro · 16 GBQ2_K_LRuns well11 GB7.543
M1 Pro · 32 GBQ3_K_MRuns great15 GB6.543
M1 Max · 32 GBQ3_K_MRuns great15 GB1273
M1 Max · 64 GBQ6_KRuns great23 GB8.673
M1 Ultra · 64 GBQ6_KRuns great23 GB15147
M1 Ultra · 128 GBQ6_KRuns great32 GB15147

Writes and reads are words a second (a token is about three quarters of a word), 4K tokens into a chat, estimated unless marked measured. The size is the one the store offers that Mac. Constants fitted October 11, 2026; how, and how wrong.

Badge for the model card

Shows on Devstral Small 2 24B's Hugging Face card which Macs run it well, linked to this page. It updates every morning.

Runs on Mac badge

Sizes and files are read from unsloth/Devstral-Small-2-24B-Instruct-2512-GGUF on Hugging Face and checked every morning. LM Arena has not rated this model, so the rating is Outmake's estimate from its size and release date; how the estimate is made. Catalog of November 4, 2026.

Your Computer

Every model is judged for this Mac. Both are under About This Mac, in the Apple menu.

Every Mac · Every graphics card

Civitai

Open in OutmakeHugging Face

New to Outmake? Download the free beta