Can I run KAT-Coder-V2.5 on a Mac?

Yes: KAT-Coder-V2.5 runs on 32 of the 54 Macs in the LLMCheck index. It needs about 24 GB at a 4K context, 26 GB at 64K. The smallest and cheapest current Mac that runs it at 15+ tok/s is the Mac mini M6 32 GB ($1,299; ~47 tok/s est.). Best measured figure: 94 tok/s on an M4 Max (community).

RAM at 4K
24
GB
RAM at 64K
26
GB, 16-bit KV
Macs that run it
32
of 54 configs
Best measured
community94
tok/s · M4 Max

KAT-Coder-V2.5 on every Mac

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 small mixture-of-experts models (40B total or less), the formula’s typical error is about 15% (the median gap to 88 measured plain-decode runs; it tends to undershoot (median −7%), and an estimate scaled from the model’s own measured run is usually closer; method).

Runs fits macOS’s default GPU limit, 15+ tok/sRuns, slowly fits, under 15 tok/sRaise GPU limit fits once the limit is raised (command below)Over GPU limit fits RAM on paper, but even a safe raise is shortDoesn't fit needs more memory than the Mac has
MacVerdictBuildSpeedMax contextModel page
MacBook Air
MacBook Air M5 16 GBDoesn't fit————
MacBook Air M5 24 GBOver GPU limit why ↓Q4_K_Mestimated44 tok/s—M5 page
MacBook Air M5 32 GBRunsQ4_K_Mestimated44 tok/s49KM5 page
MacBook Air M1 8 GB refurbDoesn't fit————
MacBook Air M4 16 GB usedDoesn't fit————
MacBook Air M2 16 GB refurbDoesn't fit————
MacBook Air M3 16 GB refurbDoesn't fit————
MacBook Air M1 16 GB refurbDoesn't fit————
MacBook Air M4 24 GB usedOver GPU limit why ↓Q4_K_Mestimated36 tok/s——
MacBook Air M2 24 GB refurbOver GPU limit why ↓Q4_K_Mestimated31 tok/s——
MacBook Air M3 24 GB refurbOver GPU limit why ↓Q4_K_Mestimated31 tok/s——
MacBook Pro
MacBook Pro M6 16 GBDoesn't fit————
MacBook Pro M5 Pro 24 GBOver GPU limit why ↓Q4_K_Mestimated72 tok/s——
MacBook Pro M6 24 GBOver GPU limit why ↓Q4_K_Mestimated47 tok/s—M6 page
MacBook Pro M6 32 GBRunsQ4_K_Mestimated47 tok/s49KM6 page
MacBook Pro M5 Pro 48 GBRunsQ4_K_Mestimated72 tok/s256K (full)—
MacBook Pro M5 Max 64 GBRunsOptiQ-4bitcommunity89 tok/s256K (full)M5 Max page
MacBook Pro M5 Pro 64 GBRunsQ4_K_Mestimated72 tok/s256K (full)—
MacBook Pro M5 Max 128 GBRunsOptiQ-4bitcommunity89 tok/s256K (full)M5 Max page
MacBook Pro M3 Max 64 GB refurbRunsQ4_K_Mestimated84 tok/s256K (full)M3 Max page
MacBook Pro M3 Max 128 GB refurbRunsQ4_K_Mestimated84 tok/s256K (full)M3 Max page
Mac mini
Mac mini M6 16 GBDoesn't fit————
Mac mini M5 Pro 24 GBOver GPU limit why ↓Q4_K_Mestimated72 tok/s——
Mac mini M6 24 GBOver GPU limit why ↓Q4_K_Mestimated47 tok/s—M6 page
Mac mini M6 32 GBRunsQ4_K_Mestimated47 tok/s49KM6 page
Mac mini M5 Pro 48 GBRunsQ4_K_Mestimated72 tok/s256K (full)—
Mac mini M5 Pro 64 GBRunsQ4_K_Mestimated72 tok/s256K (full)—
Mac mini M2 Pro 16 GB refurbDoesn't fit————
Mac mini M4 16 GB usedDoesn't fit————
Mac mini M4 Pro 24 GB usedOver GPU limit why ↓Q4_K_Mestimated66 tok/s——
Mac mini M4 24 GB usedOver GPU limit why ↓Q4_K_Mestimated36 tok/s——
Mac mini M2 Pro 32 GB refurbRaise GPU limit (macOS 15) command ↓Q4_K_Mestimated53 tok/s194K raised—
Mac mini M4 Pro 48 GB usedRunsQ4_K_Mestimated66 tok/s256K (full)—
Mac mini M4 Pro 64 GB usedRunsQ4_K_Mestimated66 tok/s256K (full)—
iMac
iMac M6 16 GBDoesn't fit————
iMac M6 24 GBOver GPU limit why ↓Q4_K_Mestimated47 tok/s—M6 page
iMac M6 32 GBRunsQ4_K_Mestimated47 tok/s49KM6 page
Mac Studio
Mac Studio M5 Max 36 GBRunsQ4_K_Mestimated89 tok/s198KM5 Max page
Mac Studio M5 Max 48 GBRunsOptiQ-4bitcommunity89 tok/s256K (full)M5 Max page
Mac Studio M5 Max 64 GBRunsOptiQ-4bitcommunity89 tok/s256K (full)M5 Max page
Mac Studio M5 Ultra 96 GBRunsQ4_K_Mestimated120 tok/s256K (full)—
Mac Studio M5 Max 128 GBRunsOptiQ-4bitcommunity89 tok/s256K (full)M5 Max page
Mac Studio M5 Ultra 256 GBRunsQ4_K_Mestimated120 tok/s256K (full)—
Mac Studio M5 Ultra 512 GBRunsQ4_K_Mestimated120 tok/s256K (full)—
Mac Studio M2 Max 32 GB refurbRaise GPU limit (macOS 15) command ↓Q4_K_Mestimated84 tok/s194K raised—
Mac Studio M1 Ultra 64 GB refurbRunsQ4_K_Mestimated99 tok/s256K (full)—
Mac Studio M4 Max 64 GB usedRuns4-bitcommunity94 tok/s256K (full)M4 Max page
Mac Studio M2 Max 64 GB refurbRunsQ4_K_Mestimated84 tok/s256K (full)—
Mac Studio M1 Ultra 128 GB refurbRunsQ4_K_Mestimated99 tok/s256K (full)—
Mac Studio M2 Ultra 128 GB refurbRunsQ4_K_Mestimated99 tok/s256K (full)—
Mac Studio M4 Max 128 GB usedRuns4-bitcommunity94 tok/s256K (full)M4 Max page
Mac Studio M2 Ultra 192 GB refurbRunsQ4_K_Mestimated99 tok/s256K (full)—
Mac Studio M3 Ultra 256 GB usedRunsQ4_K_Mestimated100 tok/s256K (full)—
Mac Studio M3 Ultra 512 GB refurbRunsQ4_K_Mestimated100 tok/s256K (full)—
Provenance key
sourced
Published figure — vendor or press review
estimated
community
Linked individual run — blog, forum or submission

Context memory is computed from its config.json (hybrid: full attention on 10 of 40 layers; Gated DeltaNet layers keep a constant state): about 1.3 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.

Raising the GPU memory limit

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 KAT-Coder-V2.5 needs about 24 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.

KAT-Coder-V2.5: memory by quantization

Weights are 35B 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. It is a mixture-of-experts model: all 35B parameters sit in memory although only 3B are read per token. The first row is the catalog’s own figure (about 24 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.

BuildWeightsRAM at 4Kat 32Kat 64KSmallest Mac
As catalogued—24 GB25 GB25 GB32 GB
Q2_K11.9 GB14 GB14 GB15 GB24 GB
Q3_K_M17.1 GB19 GB20 GB20 GB32 GB (24 GB with a raised limit)
Q4_K_M19.9 GB22 GB23 GB23 GB32 GB
Q5_K_M24.8 GB27 GB27 GB28 GB36 GB (32 GB with a raised limit)
Q6_K28.7 GB31 GB31 GB32 GB48 GB (36 GB with a raised limit)
Q8_037.1 GB39 GB40 GB40 GB64 GB (48 GB with a raised limit)
F1670.0 GB72 GB73 GB73 GB96 GB

Measured runs of KAT-Coder-V2.5

Every published (filled square) and community (hollow circle) run the index holds for KAT-Coder-V2.5: 0 published, 5 community, each linked to its source. LLMCheck runs no tests of its own. One row is one measurement session; on each chip the table above shows the lower median of the eligible runs (plain decode at 5 bits or fewer), and a binned or “not an anchor” run is shown but never used as a reference.

ChipRuntimeQuantDecodeContextDateSource
M5 Max 128 GBMLXOptiQ-4bitcommunity89.3 tok/s1k2026-08-07omlx.ai
M4 Max 64 GBMLX4-bitcommunity94 tok/s1k2026-11-07omlx.ai
M4 Max 48 GBLM Studiounstatedcommunity74.2 tok/sup to ~5.9k tokens used2026-09-19llm-bench.io
M3 Max 48 GBMLX6-bitcommunity69.5 tok/s1k2026-08-13omlx.ai
M5 32 GBMLXoQ2ecommunity49.5 tok/spp1024/tg1282026-08huggingface.co not an anchor

Recommended setup

The index has no verified Ollama tag for KAT-Coder-V2.5. 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 (32 GB), the smallest current Mac that runs it inside the default GPU limit, the index figure is ~47 tok/s (estimated) at Q4_K_M; keep the context at or under 49K there, or use a q8 KV cache.

Step-by-step: the Ollama install guide.

🛒 Which Mac to get for KAT-Coder-V2.5

Mac mini M6 (32 GB) — smallest current Mac that runs it and cheapest at 15+ tok/s, $1,299 new (Apple list price, checked 2026-11-14), ~47 tok/s (estimated). Amazon does not stock the Mac mini M6 (32 GB): Apple builds it to order.

FAQ: KAT-Coder-V2.5 on a Mac

Can I run KAT-Coder-V2.5 on a Mac?

Yes. KAT-Coder-V2.5 needs about 24 GB of unified memory at a short (4K-token) context, so it runs on 32 of the 54 Mac configurations in the LLMCheck index; 30 of them hold it inside macOS’s default GPU memory limit and 2 only after raising that limit. On 11 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 (32 GB).

How much RAM does KAT-Coder-V2.5 need?

About 24 GB at a 4K context: the weights plus about 2 GB for the runtime. Context adds KV cache: about 25 GB in total at 32K and 26 GB at 64K with the default 16-bit cache (about 25 GB at 64K with a q8 cache). These are estimates from the model’s config.json, not measurements.

How fast does KAT-Coder-V2.5 run on a Mac?

On an M5 Max, the chip the leaderboard quotes it on, the index figure is 89 tok/s (community run) at OptiQ-4bit. The best measured figure is 94 tok/s on an M4 Max 64 GB (community run, 4-bit). Every Mac’s figure is in the table on this page with its provenance: published, community run or estimated.

What is the cheapest Mac that runs KAT-Coder-V2.5 well?

Among Macs Apple sells new, the Mac mini M6 (32 GB), $1,299 new (Apple list price, checked 2026-11-14), is the cheapest that reaches 15 tok/s with KAT-Coder-V2.5, at ~47 tok/s (estimated).

Do I need to raise the GPU memory limit to run KAT-Coder-V2.5?

On 2 of the 54 configurations, yes: Mac mini M2 Pro 32 GB (used), Mac Studio M2 Max 32 GB (used). macOS lets the GPU use only part of unified memory by default (about 21 GB on the Mac mini M2 Pro (32 GB)), and KAT-Coder-V2.5 needs about 24 GB. Running sudo sysctl iogpu.wired_limit_mb=28672 in Terminal raises the limit on a 32 GB Mac until the next restart. On iMac M6 24 GB, Mac mini M5 Pro 24 GB, Mac mini M6 24 GB and 8 more even a safely raised limit is short: expect heavy swapping, a CPU split or an out-of-memory error. From 32 GB up, every Mac Apple sells new holds it inside the default limit.

Which quantization of KAT-Coder-V2.5 should I use?

Q4_K_M (about 20 GB of weights) is the common 4-bit GGUF choice. Q8_0 (about 37 GB) is closer to full quality and needs a 64 GB Mac at the default GPU limit. Q3_K_M and Q2_K shrink it to about 17 and 12 GB at a quality cost. The index’s reference figure (89 tok/s on an M5 Max) is for its OptiQ-4bit build.

How much context can KAT-Coder-V2.5 use on a Mac?

Its own window is 256K tokens. At macOS’s default GPU limit, the Mac mini M6 (32 GB) holds about 49K. The full window needs about 30 GB, which a 48 GB Mac holds at the default limit. A q8 KV cache roughly doubles these lengths.

Cite this page

LLMCheck. (2026). Can I run KAT-Coder-V2.5 on a Mac? llmcheck.net. https://llmcheck.net/can-i-run/kat-coder-v25/ (updated ). CC BY 4.0.

How to cite LLMCheck →Data: CC BY 4.0Markdown version

Related

Figures from the LLMCheck index, data version 2026-11-18. Verdicts and context limits are estimates from each model’s catalog memory figure and config.json.