Yes: Devstral Small 24B runs on 43 of 54 Macs in the LLMCheck index. It needs about 14 GB of unified memory at a 4K context and about 25 GB at 64K. Smallest current Mac: the Mac mini M6 24 GB; cheapest at 15+ tok/s: the Mac mini M5 Pro 24 GB ($1,699; ~17 tok/s est.). No measured run yet.
One row per Mac configuration, grouped by product line. The build is the one each speed is for; the context column is the longest prompt that fits inside macOS’s default GPU memory limit with a 16-bit KV cache (“raised” where the limit has to be raised first; “full” when the whole window fits). Speeds are decode tok/s from the LLMCheck index, each with its provenance mark. For dense models, the formula’s typical error is about 9% (the median gap to 75 measured plain-decode runs; it tends to undershoot (median −6%), and an estimate scaled from the model’s own measured run is usually closer; method).
| Mac | Verdict | Build | Speed | Max context | Model page |
|---|---|---|---|---|---|
| MacBook Air | |||||
| MacBook Air M5 16 GB | Over GPU limit why ↓ | Q4_K_M | estimated8 tok/s | — | — |
| MacBook Air M5 24 GB | Runs, slowly | Q4_K_M | estimated8 tok/s | 26K | — |
| MacBook Air M5 32 GB | Runs, slowly | Q4_K_M | estimated8 tok/s | 69K | — |
| MacBook Air M1 8 GB refurb | Doesn't fit | — | — | — | — |
| MacBook Air M4 16 GB used | Over GPU limit why ↓ | Q4_K_M | estimated7 tok/s | — | — |
| MacBook Air M2 16 GB refurb | Over GPU limit why ↓ | Q4_K_M | estimated6 tok/s | — | — |
| MacBook Air M3 16 GB refurb | Over GPU limit why ↓ | Q4_K_M | estimated6 tok/s | — | — |
| MacBook Air M1 16 GB refurb | Over GPU limit why ↓ | Q4_K_M | estimated4 tok/s | — | — |
| MacBook Air M4 24 GB used | Runs, slowly | Q4_K_M | estimated7 tok/s | 15K | — |
| MacBook Air M2 24 GB refurb | Runs, slowly | Q4_K_M | estimated6 tok/s | 15K | — |
| MacBook Air M3 24 GB refurb | Runs, slowly | Q4_K_M | estimated6 tok/s | 15K | — |
| MacBook Pro | |||||
| MacBook Pro M6 16 GB | Over GPU limit why ↓ | Q4_K_M | estimated8 tok/s | — | — |
| MacBook Pro M5 Pro 24 GB | Runs | Q4_K_M | estimated17 tok/s | 26K | — |
| MacBook Pro M6 24 GB | Runs, slowly | Q4_K_M | estimated9 tok/s | 26K | — |
| MacBook Pro M6 32 GB | Runs, slowly | Q4_K_M | estimated9 tok/s | 69K | — |
| MacBook Pro M5 Pro 48 GB | Runs | Q4_K_M | estimated17 tok/s | 143K | — |
| MacBook Pro M5 Max 64 GB | Runs | Q4_K_M | estimated33 tok/s | 229K | — |
| MacBook Pro M5 Pro 64 GB | Runs | Q4_K_M | estimated17 tok/s | 229K | — |
| MacBook Pro M5 Max 128 GB | Runs | Q4_K_M | estimated33 tok/s | 256K (full) | — |
| MacBook Pro M3 Max 64 GB refurb | Runs | Q4_K_M | estimated22 tok/s | 206K | — |
| MacBook Pro M3 Max 128 GB refurb | Runs | Q4_K_M | estimated22 tok/s | 256K (full) | — |
| Mac mini | |||||
| Mac mini M6 16 GB | Over GPU limit why ↓ | Q4_K_M | estimated8 tok/s | — | — |
| Mac mini M5 Pro 24 GB | Runs | Q4_K_M | estimated17 tok/s | 26K | — |
| Mac mini M6 24 GB | Runs, slowly | Q4_K_M | estimated9 tok/s | 26K | — |
| Mac mini M6 32 GB | Runs, slowly | Q4_K_M | estimated9 tok/s | 69K | — |
| Mac mini M5 Pro 48 GB | Runs | Q4_K_M | estimated17 tok/s | 143K | — |
| Mac mini M5 Pro 64 GB | Runs | Q4_K_M | estimated17 tok/s | 229K | — |
| Mac mini M2 Pro 16 GB refurb | Over GPU limit why ↓ | Q4_K_M | estimated11 tok/s | — | — |
| Mac mini M4 16 GB used | Over GPU limit why ↓ | Q4_K_M | estimated7 tok/s | — | — |
| Mac mini M4 Pro 24 GB used | Runs | Q4_K_M | estimated15 tok/s | 15K | — |
| Mac mini M4 24 GB used | Runs, slowly | Q4_K_M | estimated7 tok/s | 15K | — |
| Mac mini M2 Pro 32 GB refurb | Runs, slowly | Q4_K_M | estimated11 tok/s | 47K | — |
| Mac mini M4 Pro 48 GB used | Runs | Q4_K_M | estimated15 tok/s | 135K | — |
| Mac mini M4 Pro 64 GB used | Runs | Q4_K_M | estimated15 tok/s | 206K | — |
| iMac | |||||
| iMac M6 16 GB | Over GPU limit why ↓ | Q4_K_M | estimated8 tok/s | — | — |
| iMac M6 24 GB | Runs, slowly | Q4_K_M | estimated9 tok/s | 26K | — |
| iMac M6 32 GB | Runs, slowly | Q4_K_M | estimated9 tok/s | 69K | — |
| Mac Studio | |||||
| Mac Studio M5 Max 36 GB | Runs | Q4_K_M | estimated25 tok/s | 87K | — |
| Mac Studio M5 Max 48 GB | Runs | Q4_K_M | estimated33 tok/s | 143K | — |
| Mac Studio M5 Max 64 GB | Runs | Q4_K_M | estimated33 tok/s | 229K | — |
| Mac Studio M5 Ultra 96 GB | Runs | Q4_K_M | estimated63 tok/s | 256K (full) | — |
| Mac Studio M5 Max 128 GB | Runs | Q4_K_M | estimated33 tok/s | 256K (full) | — |
| Mac Studio M5 Ultra 256 GB | Runs | Q4_K_M | estimated63 tok/s | 256K (full) | — |
| Mac Studio M5 Ultra 512 GB | Runs | Q4_K_M | estimated63 tok/s | 256K (full) | — |
| Mac Studio M2 Max 32 GB refurb | Runs | Q4_K_M | estimated22 tok/s | 47K | — |
| Mac Studio M1 Ultra 64 GB refurb | Runs | Q4_K_M | estimated42 tok/s | 206K | — |
| Mac Studio M4 Max 64 GB used | Runs | Q4_K_M | estimated30 tok/s | 206K | — |
| Mac Studio M2 Max 64 GB refurb | Runs | Q4_K_M | estimated22 tok/s | 206K | — |
| Mac Studio M1 Ultra 128 GB refurb | Runs | Q4_K_M | estimated42 tok/s | 256K (full) | — |
| Mac Studio M2 Ultra 128 GB refurb | Runs | Q4_K_M | estimated42 tok/s | 256K (full) | — |
| Mac Studio M4 Max 128 GB used | Runs | Q4_K_M | estimated30 tok/s | 256K (full) | — |
| Mac Studio M2 Ultra 192 GB refurb | Runs | Q4_K_M | estimated42 tok/s | 256K (full) | — |
| Mac Studio M3 Ultra 256 GB used | Runs | Q4_K_M | estimated43 tok/s | 256K (full) | — |
| Mac Studio M3 Ultra 512 GB refurb | Runs | Q4_K_M | estimated43 tok/s | 256K (full) | — |
Context memory is computed from its config.json (GQA on all 40 layers (8 kv heads x 128)): about 10.7 GB of 16-bit KV cache at 64K tokens. A q8 cache (llama.cpp -ctk q8_0 -ctv q8_0, Ollama OLLAMA_KV_CACHE_TYPE=q8_0) halves that part. Runtimes also grow a compute buffer with context, so a fit within about 1 GB of the limit is optimistic.
macOS lets the GPU wire only part of unified memory by default (Metal’s recommended working set; M5 and M6
Macs run macOS 26 or later, whose default is higher), and Devstral Small 24B needs about 14 GB. On the Macs marked
Raise GPU limit it fits in RAM but not inside that default. The default
is a performance threshold, not a hard wall: past it, the runtime moves part of it to the CPU or shortens the context (llama.cpp, Ollama), runs it slowly (MLX), or stops with an out-of-memory error. One command raises the limit until the next restart
(sudo sysctl iogpu.wired_limit_mb=0 restores the default sooner); it leaves macOS a reserve, and other
apps get less memory while the model is loaded. On the Macs marked Over GPU limit even that raise falls short: the model fits the Mac’s RAM on paper, but a smaller quant or a bigger Mac is the realistic route.
sudo sysctl iogpu.wired_limit_mb=12288, 12 GB) is short of the ~14 GB it needs: expect heavy swapping, a CPU split or an out-of-memory errorsudo sysctl iogpu.wired_limit_mb=12288, 12 GB) is short of the ~14 GB it needs: expect heavy swapping, a CPU split or an out-of-memory errorWeights are 24B parameters at each format’s effective size (the index’s GB-per-billion table), plus about 2 GB for the runtime and a short context; longer contexts add the KV cache computed from the model’s config.json at 16 bits. The first row is the catalog’s own figure (about 14 GB at a 4K context), the one every verdict on this page uses. “Smallest Mac” is the smallest memory size among the Macs Apple sells new that holds it at a 4K context inside that Mac’s default GPU limit. Estimates, not measurements.
| Build | Weights | RAM at 4K | at 32K | at 64K | Smallest Mac |
|---|---|---|---|---|---|
| As catalogued | — | 14 GB | 19 GB | 24 GB | 24 GB |
| Q2_K | 8.2 GB | 10 GB | 15 GB | 20 GB | 16 GB |
| Q3_K_M | 11.8 GB | 14 GB | 18 GB | 24 GB | 24 GB |
| Q4_K_M | 13.7 GB | 16 GB | 20 GB | 26 GB | 24 GB |
| Q5_K_M | 17.0 GB | 19 GB | 24 GB | 29 GB | 32 GB |
| Q6_K | 19.7 GB | 22 GB | 26 GB | 32 GB | 32 GB |
| Q8_0 | 25.4 GB | 27 GB | 32 GB | 38 GB | 36 GB |
| F16 | 48.0 GB | 50 GB | 55 GB | 60 GB | 64 GB |
The index holds no published or community run of Devstral Small 24B yet, so every speed on this page is an estimate from the memory-bandwidth model. LLMCheck runs no tests of its own: if you have run it on a Mac, submit the figure with its setup →
The index has no verified Ollama tag for Devstral Small 24B. Start from its listing and copy the exact name of a build that fits: Search the Ollama library → · model card →
GGUF builds run in LM Studio or llama.cpp; MLX builds run in LM Studio or mlx-lm.
On the Mac mini M6 (24 GB), the smallest current Mac that runs it inside the default GPU limit, the index figure is ~9 tok/s (estimated) at Q4_K_M; keep the context at or under 26K there, or use a q8 KV cache.
Step-by-step: the Ollama install guide.
Mac mini M6 (24 GB) — smallest current Mac that runs it, $1,099 new (Apple list price, checked 2026-11-14), ~9 tok/s (estimated). Mac mini M5 Pro (24 GB) — cheapest at 15+ tok/s, $1,699 new (Apple list price, checked 2026-11-14), ~17 tok/s (estimated).
As an Amazon Associate, LLMCheck earns from qualifying purchases. Affiliate links cost you nothing extra and never influence the verdicts.
Yes. Devstral Small 24B needs about 14 GB of unified memory at a short (4K-token) context, so it runs on 43 of the 54 Mac configurations in the LLMCheck index. On 10 more it fits the RAM, but not even a safely raised GPU limit. The smallest Mac Apple sells new that runs it is the Mac mini M6 (24 GB).
About 14 GB at a 4K context: the weights plus about 2 GB for the runtime. Context adds KV cache: about 19 GB in total at 32K and 25 GB at 64K with the default 16-bit cache (about 20 GB at 64K with a q8 cache). These are estimates from the model’s config.json, not measurements.
On an M5 Max, the chip the leaderboard quotes it on, the index figure is ~33 tok/s (estimated) at Q4_K_M. No published or community run is in the index yet; every speed is an estimate from the memory-bandwidth model. Every Mac’s figure is in the table on this page with its provenance: published, community run or estimated.
Among Macs Apple sells new, the Mac mini M5 Pro (24 GB), $1,699 new (Apple list price, checked 2026-11-14), is the cheapest that reaches 15 tok/s with Devstral Small 24B, at ~17 tok/s (estimated).
On iMac M6 16 GB, Mac mini M6 16 GB, MacBook Air M5 16 GB and 7 more even a safely raised limit is short: expect heavy swapping, a CPU split or an out-of-memory error. From 24 GB up, every Mac Apple sells new holds it inside the default limit.
Q4_K_M (about 14 GB of weights) is the common 4-bit GGUF choice. Q8_0 (about 25 GB) is closer to full quality and needs a 36 GB Mac at the default GPU limit. Q3_K_M and Q2_K shrink it to about 12 and 8 GB at a quality cost. The index’s reference figure (33 tok/s on an M5 Max) is for its Q4_K_M build.
Its own window is 256K tokens. At macOS’s default GPU limit, the Mac mini M6 (24 GB) holds about 26K. The full window needs about 57 GB, which a 96 GB Mac holds at the default limit. A q8 KV cache roughly doubles these lengths.
LLMCheck. (2026). Can I run Devstral Small 24B on a Mac? llmcheck.net. https://llmcheck.net/can-i-run/devstral-small-24b/ (updated ). CC BY 4.0.