Local LLM on the MacBook Air M5: best models and tokens/s

The best local LLM for the MacBook Air M5, by memory

Memory	Strongest that runs comfortably	Writes	Suggested by Twoody by default
16 GB	Gemma 4 12B One click in Twoody	14–20 tokens/s	Qwen3.5 9B (7.4 GB)
24 GB	Qwen3 14B One click in Twoody	11–16 tokens/s	Qwen3.5 9B (7.4 GB)
32 GB	Qwen3.6 35B-A3B One click in Twoody	39–77 tokens/s	Qwen3.5 9B (7.9 GB)

Speeds are estimates, not measurements made with Twoody, computed from public llama.cpp benchmarks. Figures reviewed on October 16, 2026.

“Strongest that runs comfortably”: the strongest model that fits comfortably in this memory and writes at least at your reading speed (5 tokens per second). “Suggested by Twoody by default”: the model the app installs first on this Mac, and the memory it needs there.

Twoody is in private beta: leave your email to request access.

Every model that runs well on the MacBook Air M5 with 16 GB

31 of the 62 models the calculator knows (Twoody's catalogue and a few larger ones) run well here, the strongest first: the memory Twoody's app says each one needs on this computer, how fast it writes, and how long it takes to read a document.

Model	Memory	Writes	Reads a document
Gemma 4 12B One click in Twoody	Needs 7.6 GB of memory Runs comfortably	14–20 tokens/s	20 pages in 39 s
GLM-4.6V Flash One click in Twoody	Needs 7.9 GB of memory Runs comfortably	16–23 tokens/s	20 pages in 34 s
Qwen3.5 9B One click in Twoody	Needs 7.4 GB of memory Runs comfortably	12–24 tokens/s	20 pages in 32 s
GLM-4 9B One click in Twoody	Needs 6.7 GB of memory Runs comfortably	16–23 tokens/s	20 pages in 31 s
Ministral 3 8B One click in Twoody	Needs 8.4 GB of memory Runs comfortably	19–27 tokens/s	20 pages in 29 s
Granite 4.1 8B One click in Twoody	Needs 8 GB of memory Runs comfortably	17–25 tokens/s	20 pages in 29 s
Qwen3 8B One click in Twoody	Needs 7.2 GB of memory Runs comfortably	19–27 tokens/s	20 pages in 27 s
Llama 3.1 8B One click in Twoody	Needs 6.9 GB of memory Runs comfortably	20–28 tokens/s	20 pages in 26 s
Olmo 3 7B One click in Twoody	Needs 8.2 GB of memory Runs comfortably	20–28 tokens/s	20 pages in 24 s
Gemma 4 E4B One click in Twoody	Needs 6.6 GB of memory Runs comfortably	32–46 tokens/s	20 pages in 15 s
Qwen3.5 4B One click in Twoody	Needs 4.4 GB of memory Runs comfortably	23–47 tokens/s	20 pages in 16 s
Qwen3 4B Instruct 2507 One click in Twoody	Needs 4.9 GB of memory Runs comfortably	35–51 tokens/s	20 pages in 13 s
Nemotron 3 Nano 4B One click in Twoody	Needs 3.3 GB of memory Runs comfortably	23–45 tokens/s	20 pages in 13 s
Ministral 3 3B One click in Twoody	Needs 5 GB of memory Runs comfortably	42–61 tokens/s	20 pages in 12 s
Phi-4 mini One click in Twoody	Needs 4.6 GB of memory Runs comfortably	36–52 tokens/s	20 pages in 12 s
Llama 3.2 3B One click in Twoody	Needs 3.9 GB of memory Runs comfortably	44–63 tokens/s	20 pages in 10 s
Gemma 4 E2B One click in Twoody	Needs 4.7 GB of memory Runs comfortably	58–83 tokens/s	20 pages in 8 s
Qwen3.5 2B One click in Twoody	Needs 2.7 GB of memory Runs comfortably	45–90 tokens/s	20 pages in 8 s
Qwen3.5 0.8B One click in Twoody	Needs 1.5 GB of memory Runs comfortably	71–213 tokens/s	20 pages in 3 s
DeepSeek-Coder-V2 Lite One click in Twoody	Needs 10.4 GB of memory Runs — close other apps	49–71 tokens/s	20 pages in 10 s
Ministral 3 14B Reasoning One click in Twoody	Needs 10.5 GB of memory Runs — close other apps	12–17 tokens/s	20 pages in 46 s
GLM-Z1 9B One click in Twoody	Needs 6.7 GB of memory Runs comfortably	16–23 tokens/s	20 pages in 31 s
DeepSeek-R1 8B One click in Twoody	Needs 7.2 GB of memory Runs comfortably	19–27 tokens/s	20 pages in 27 s
DeepSeek-R1 Distill Llama 8B One click in Twoody	Needs 6.9 GB of memory Runs comfortably	20–28 tokens/s	20 pages in 26 s
DeepSeek-R1 Distill Qwen 7B One click in Twoody	Needs 5.5 GB of memory Runs comfortably	21–30 tokens/s	20 pages in 25 s
Qwen3 4B Thinking 2507 One click in Twoody	Needs 4.9 GB of memory Runs comfortably	35–51 tokens/s	20 pages in 13 s
Phi-4 mini reasoning One click in Twoody	Needs 4.6 GB of memory Runs comfortably	36–52 tokens/s	20 pages in 12 s
SmolLM3 3B One click in Twoody	Needs 3.2 GB of memory Runs comfortably	46–66 tokens/s	20 pages in 10 s
LFM2.5 2.6B One click in Twoody	Needs 2.1 GB of memory Runs comfortably	37–74 tokens/s	20 pages in 9 s
MiniCPM5 2B One click in Twoody	Needs 2.5 GB of memory Runs comfortably	52–75 tokens/s	20 pages in 8 s
Llama 3.2 1B One click in Twoody	Needs 1.6 GB of memory Runs comfortably	67–201 tokens/s	20 pages in 4 s

Besides these, 6 models would run partly on the processor, much slower, and 25 are too big for this computer. See them all in the calculator.

Each model's memory is what Twoody shows in “Models” for this computer: the same formula, the same files (the one for reading images included), the same conversation length — catalogue checked on October 15, 2026.

MacBook Air M5 compared with MacBook Air M4, MacBook Air M3, MacBook Air M2 and MacBook Air M1

Computer	Memory	Qwen3.5 9B writes	Reads a document	Strongest that runs comfortably
MacBook Air M5	16 GB	12–24 tokens/s	20 pages in 32 s	Gemma 4 12B 14–20 tokens/s
MacBook Air M4	16 GB	9.3–19 tokens/s	20 pages in 2 min	Gemma 4 12B 11–15 tokens/s
MacBook Air M3	16 GB	8–16 tokens/s	20 pages in 2 min	Gemma 4 12B 9.4–14 tokens/s
MacBook Air M2	16 GB	8.1–16 tokens/s	20 pages in 2 min	Gemma 4 12B 9.5–14 tokens/s
MacBook Air M1	16 GB	5.4–11 tokens/s	20 pages in 3 min	Gemma 4 12B 6.3–9 tokens/s

Memory decides which models fit; the chip, how fast they write. Each Mac is shown with 16 GB, or the nearest size it comes with.

This Mac comes with 16, 24 or 32 GB of memory, shared between its processor and its graphics, which may use most of it: that memory decides which models fit, and it cannot be added later.

It reads its memory at 153 GB/s. Speed follows that bandwidth more than the processor: that is why a Pro, Max or Ultra chip writes faster with the same model.

Twoody is in private beta on Macs with Apple silicon with macOS 13.3 or later: it suggests the model that suits this Mac, installs it in one click and runs it with llama.cpp; with macOS 14, it can also install MLX, Apple's engine.

Questions about running an LLM locally

Which local LLM for the MacBook Air M5 with 16 GB?

The strongest model that runs comfortably there is Gemma 4 12B, which writes 14–20 tokens/s. It installs in one click from Twoody's catalogue. Twoody suggests Qwen3.5 9B by default, which needs 7.4 GB of memory there: one of the models it has measured with its memory, its tools and your documents.

How much memory does the MacBook Air M5 need for a local LLM?

The strongest models that run comfortably there, by memory: with 16 GB, Gemma 4 12B (14–20 tokens/s, one click in Twoody); with 24 GB, Qwen3 14B (11–16 tokens/s, one click in Twoody); with 32 GB, Qwen3.6 35B-A3B (39–77 tokens/s, one click in Twoody). A Mac's memory cannot be added later: choose it for the models you want to run.

How many tokens per second does the MacBook Air M5 write?

With Qwen3.5 9B, the model Twoody suggests there, it writes 12–24 tokens/s and reads a document of about 20 pages in 32 s. These speeds are estimates from public llama.cpp benchmarks, not measurements made with Twoody.

How does the MacBook Air M5 compare with the MacBook Air M4, MacBook Air M3, MacBook Air M2 and MacBook Air M1?

With Qwen3.5 9B, the model Twoody suggests: MacBook Air M5, 12–24 tokens/s; MacBook Air M4, 9.3–19 tokens/s; MacBook Air M3, 8–16 tokens/s; MacBook Air M2, 8.1–16 tokens/s; MacBook Air M1, 5.4–11 tokens/s. Memory decides which models fit; the chip, how fast they write.

Does Twoody run on the MacBook Air M5?

Yes, in the private beta, with macOS 13.3 or later: Twoody suggests the model that suits it, installs it in one click and runs it on the Mac's graphics with llama.cpp. Leave your email to join the waitlist.

Other computers and guides

Try Twoody before it opens to everyone.

The private beta is free, on Mac, Windows and Linux, and on iPhone and Android with Twoody on your computer. Leave your email to ask for an invitation: we will write to you when it is ready for you.

Request beta access

A local LLM on the MacBook Air M5