Can I run Recursive Flint on a Mac?

Yes: Recursive Flint runs on 43 of 54 Macs in the LLMCheck index. It needs about 20 GB at a 4K context, 22 GB at 64K. The smallest and cheapest current Mac at 15+ tok/s is the Mac mini M6 24 GB ($1,099; ~71 tok/s est.), with the GPU limit raised. Best measured: 132.5 tok/s on an M5 Max (community).

RAM at 4K
20
GB
RAM at 64K
22
GB, 16-bit KV
Macs that run it
43
of 54 configs
Best measured
community132.5
tok/s · M5 Max

Recursive Flint 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)Doesn'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 GBRaise GPU limit command ↓3.8 bpwestimated67 tok/s4K raised—
MacBook Air M5 32 GBRunsQ4_K_Mestimated57 tok/s172K—
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 usedRaise GPU limit command ↓3.8 bpwestimated56 tok/s4K raised—
MacBook Air M2 24 GB refurbRaise GPU limit command ↓3.8 bpwestimated49 tok/s4K raised—
MacBook Air M3 24 GB refurbRaise GPU limit command ↓3.8 bpwestimated49 tok/s4K raised—
MacBook Pro
MacBook Pro M6 16 GBDoesn't fit————
MacBook Pro M5 Pro 24 GBRaise GPU limit command ↓3.8 bpwestimated92 tok/s4K raisedM5 Pro page
MacBook Pro M6 24 GBRaise GPU limit command ↓3.8 bpwestimated71 tok/s4K raisedM6 page
MacBook Pro M6 32 GBRunsQ4_K_Mestimated61 tok/s172KM6 page
MacBook Pro M5 Pro 48 GBRunsQ4_1community92 tok/s256K (full)M5 Pro page
MacBook Pro M5 Max 64 GBRuns4-bitcommunity123 tok/s256K (full)M5 Max page
MacBook Pro M5 Pro 64 GBRunsQ4_1community92 tok/s256K (full)M5 Pro page
MacBook Pro M5 Max 128 GBRuns4-bitcommunity123 tok/s256K (full)M5 Max page
MacBook Pro M3 Max 64 GB refurbRunsQ4_K_Mestimated89 tok/s256K (full)—
MacBook Pro M3 Max 128 GB refurbRunsQ4_K_Mestimated89 tok/s256K (full)—
Mac mini
Mac mini M6 16 GBDoesn't fit————
Mac mini M5 Pro 24 GBRaise GPU limit command ↓3.8 bpwestimated92 tok/s4K raisedM5 Pro page
Mac mini M6 24 GBRaise GPU limit command ↓3.8 bpwestimated71 tok/s4K raisedM6 page
Mac mini M6 32 GBRunsQ4_K_Mestimated61 tok/s172KM6 page
Mac mini M5 Pro 48 GBRunsQ4_1community92 tok/s256K (full)M5 Pro page
Mac mini M5 Pro 64 GBRunsQ4_1community92 tok/s256K (full)M5 Pro page
Mac mini M2 Pro 16 GB refurbDoesn't fit————
Mac mini M4 16 GB usedDoesn't fit————
Mac mini M4 Pro 24 GB usedRaise GPU limit command ↓3.8 bpwestimated80 tok/s4K raisedM4 Pro page
Mac mini M4 24 GB usedRaise GPU limit command ↓3.8 bpwestimated56 tok/s4K raised—
Mac mini M2 Pro 32 GB refurbRunsQ4_K_Mestimated68 tok/s49K—
Mac mini M4 Pro 48 GB usedRunsQ4_1community80 tok/s256K (full)M4 Pro page
Mac mini M4 Pro 64 GB usedRunsQ4_1community80 tok/s256K (full)M4 Pro page
iMac
iMac M6 16 GBDoesn't fit————
iMac M6 24 GBRaise GPU limit command ↓3.8 bpwestimated71 tok/s4K raisedM6 page
iMac M6 32 GBRunsQ4_K_Mestimated61 tok/s172KM6 page
Mac Studio
Mac Studio M5 Max 36 GBRunsINT4community131 tok/s256K (full)M5 Max page
Mac Studio M5 Max 48 GBRuns4-bitcommunity123 tok/s256K (full)M5 Max page
Mac Studio M5 Max 64 GBRuns4-bitcommunity123 tok/s256K (full)M5 Max page
Mac Studio M5 Ultra 96 GBRunsQ4_K_Mestimated131 tok/s256K (full)—
Mac Studio M5 Max 128 GBRuns4-bitcommunity123 tok/s256K (full)M5 Max page
Mac Studio M5 Ultra 256 GBRunsQ4_K_Mestimated131 tok/s256K (full)—
Mac Studio M5 Ultra 512 GBRunsQ4_K_Mestimated131 tok/s256K (full)—
Mac Studio M2 Max 32 GB refurbRunsQ4_K_Mestimated89 tok/s49K—
Mac Studio M1 Ultra 64 GB refurbRunsQ4_K_Mestimated89 tok/s256K (full)—
Mac Studio M4 Max 64 GB usedRunsQ4_K_Mestimated89 tok/s256K (full)—
Mac Studio M2 Max 64 GB refurbRunsQ4_K_Mestimated89 tok/s256K (full)—
Mac Studio M1 Ultra 128 GB refurbRunsQ4_K_Mestimated89 tok/s256K (full)—
Mac Studio M2 Ultra 128 GB refurbRunsQ4_K_Mestimated89 tok/s256K (full)—
Mac Studio M4 Max 128 GB usedRunsQ4_K_Mestimated89 tok/s256K (full)—
Mac Studio M2 Ultra 192 GB refurbRunsQ4_K_Mestimated89 tok/s256K (full)—
Mac Studio M3 Ultra 256 GB usedRuns4-bitcommunity91 tok/s256K (full)M3 Ultra page
Mac Studio M3 Ultra 512 GB refurbRuns4-bitcommunity91 tok/s256K (full)M3 Ultra page
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 (14 full layers + 35 sliding-window layers (window 513)): about 1.9 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 Recursive Flint needs about 20 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.

Recursive Flint: memory by quantization

Weights are 39.8B 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 39.8B parameters sit in memory although only 2.5B are read per token. The first row is the catalog’s own figure (about 20 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—20 GB21 GB22 GB32 GB (24 GB with a raised limit)
Q2_K13.5 GB16 GB16 GB17 GB24 GB
Q3_K_M19.5 GB22 GB22 GB23 GB32 GB
Q4_K_M22.7 GB25 GB26 GB26 GB32 GB
Q5_K_M28.3 GB30 GB31 GB32 GB48 GB (36 GB with a raised limit)
Q6_K32.6 GB35 GB35 GB36 GB48 GB
Q8_042.2 GB44 GB45 GB46 GB64 GB
F1679.6 GB82 GB82 GB83 GB128 GB (96 GB with a raised limit)

Measured runs of Recursive Flint

Every published (filled square) and community (hollow circle) run the index holds for Recursive Flint: 2 published, 18 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
M3 Ultra —MLXINT4community104.5 tok/s19-token prompt, 512 generated, median of 52026-10-21x.com not an anchor
M3 Ultra —MLX4-bitcommunity90.6 tok/s1K tokens per request, batch 12026-10-21x.com
M3 Ultra 512 GBMLXBF16community51.3 tok/s1K, single request2026-10-21x.com
M5 Max 64 GBMLXmixed 3/4-bitcommunity163.5 tok/s1k2026-11-08omlx.ai not an anchor
M5 Max 64 GBMLXmixed 3/4-bitcommunity147.2 tok/s1k2026-11-10omlx.ai not an anchor
M5 Max 64 GBMLXoQ4ecommunity132.5 tok/s1k2026-10-29omlx.ai
M5 Max 36 GBMLXINT4community131 tok/s~1K context, greedy, 256 generated2026-10-22github.com binned
M5 Max 128 GBllama.cppQ4_1community129.6 tok/s512-token coding answer at T 1.0 after a short prompt2026-10-26github.com
M5 Max 128 GBMLX4-bitcommunity123.4 tok/s1k2026-11-01omlx.ai
M5 Max 128 GBMLXOptiQ-4bitcommunity115.2 tok/s1k2026-10-25omlx.ai
M5 Max 128 GBMLXoQ6ecommunity110.2 tok/s1k2026-10-22omlx.ai not an anchor
M5 Max 128 GBMLXMXFP4community77.4 tok/s1k2026-10-29omlx.ai
M5 Pro 64 GBOllama4-bitcommunity115 tok/s18-token prompt, 1,075 generated2026-10-21x.com
M5 Pro 48 GBllama.cppQ4_1community92.36 tok/s512-token coding answers, median of 5 at T 1.02026-10-22github.com
M5 Pro 48 GBMLXoQ4ecommunity80.7 tok/s1k2026-10-22omlx.ai
M4 Pro 48 GBrecINT4sourced131.2 tok/s1K prompt, 1,024 generated, 5 runs2026-10-21github.com not an anchor
M4 Pro 48 GBllama.cpp4-bitsourced88 tok/s1K prompt, 1,024 generated, 5 runs2026-10-21github.com not an anchor
M4 Pro 48 GBllama.cppQ4_1community79.57 tok/s512-token coding answers, median of 5 at T 1.02026-10-22github.com
M6 32 GBMLXmixed 3/4-bitcommunity83.5 tok/s1k2026-11-08omlx.ai not an anchor
M3 Pro 36 GBllama.cppQ4_1community50.87 tok/s512-token coding answers at T 1.0, -c 327682026-10-22github.com

Recommended setup

The quickest start is Ollama. Pull the model by its tag (checked on ollama.com, 2026-11-14; the file is about 23 GB):

ollama run flint:40b

Prefer a window to a terminal? LM Studio downloads the same model with one click. · model card →

On the Mac mini M6 (32 GB), the smallest current Mac that runs it inside the default GPU limit, the index figure is ~61 tok/s (estimated) at Q4_K_M; keep the context at or under 172K there, or use a q8 KV cache. The flint:40b file fits there.

Step-by-step: the Ollama install guide.

🛒 Which Mac to get for Recursive Flint

Mac mini M6 (24 GB) — smallest current Mac that runs it and cheapest at 15+ tok/s, $1,099 new (Apple list price, checked 2026-11-14), ~71 tok/s (estimated) once the GPU memory limit is raised. For Ollama: the flint:40b tag is about 23 GB, so it needs a 32 GB Mac; on 24 GB use a ~3-bit build after raising the GPU limit.

As an Amazon Associate, LLMCheck earns from qualifying purchases. Affiliate links cost you nothing extra and never influence the verdicts.

FAQ: Recursive Flint on a Mac

Can I run Recursive Flint on a Mac?

Yes. Recursive Flint needs about 20 GB of unified memory at a short (4K-token) context, so it runs on 43 of the 54 Mac configurations in the LLMCheck index; 32 of them hold it inside macOS’s default GPU memory limit and 11 only after raising that limit. The smallest Mac Apple sells new that runs it is the Mac mini M6 (24 GB).

How much RAM does Recursive Flint need?

About 20 GB at a 4K context: the weights plus about 2 GB for the runtime. Context adds KV cache: about 21 GB in total at 32K and 22 GB at 64K with the default 16-bit cache (about 21 GB at 64K with a q8 cache). These are estimates from the model’s config.json, not measurements.

How fast does Recursive Flint run on a Mac?

On an M5 Max, the chip the leaderboard quotes it on, the index figure is 123 tok/s (community run) at 4-bit. The best measured figure is 132.5 tok/s on an M5 Max 64 GB (community run, oQ4e). 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 Recursive Flint well?

Among Macs Apple sells new, the Mac mini M6 (24 GB), $1,099 new (Apple list price, checked 2026-11-14), is the cheapest that reaches 15 tok/s with Recursive Flint, at ~71 tok/s (estimated), after raising the GPU memory limit.

Do I need to raise the GPU memory limit to run Recursive Flint?

On 11 of the 54 configurations, yes: iMac M6 24 GB, Mac mini M5 Pro 24 GB, Mac mini M6 24 GB and 8 more. macOS lets the GPU use only part of unified memory by default (about 18 GB on the iMac M6 (24 GB)), and Recursive Flint needs about 20 GB. Running sudo sysctl iogpu.wired_limit_mb=20480 in Terminal raises the limit on a 24 GB Mac until the next restart. From 32 GB up, every Mac Apple sells new holds it inside the default limit.

Which quantization of Recursive Flint should I use?

Q4_K_M (about 23 GB of weights) is the common 4-bit GGUF choice. Q8_0 (about 42 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 20 and 14 GB at a quality cost. The index’s reference figure (123 tok/s on an M5 Max) is for its 4-bit build.

How much context can Recursive Flint use on a Mac?

Its own window is 256K tokens. At macOS’s default GPU limit, the Mac mini M6 (32 GB) holds about 172K. The full window needs about 28 GB, which a 36 GB Mac holds at the default limit. A q8 KV cache roughly doubles these lengths.

Cite this page

LLMCheck. (2026). Can I run Recursive Flint on a Mac? llmcheck.net. https://llmcheck.net/can-i-run/recursive-flint/ (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.