What is the best LLM for a Mac with 32 GB?

Of the models with a published Intelligence Index, Qwen3.8 27B scores highest among those that run well in 32 GB (4-bit). The biggest that runs well is Qwen3.6 35B A3B. For code: Qwen3 Coder 30B A3B Instruct.

Is 32 GB enough for local AI on a Mac?

38 of the 73 open models here run well in it at 4 bits or better. macOS lets the GPU use about 24 GB of it by default; the model and its context have to fit in that.

Can a 32 GB Mac run gpt-oss-20b?

Yes, native 4.25-bit, at about 28–51 tokens a second on an M4 and 95–168 on an M1 Max.

Best local LLMs for a 32 GB Mac (M1 to M6): what runs, how fast

Macs with 32 GB

7 chips from M1 to M6 come with 32 GB. Open one for its own list and speeds.

Runs well 38

Fits with room for context, at 10 or more tokens a second on the M1 Max.

Model	Version	Memory	M4	M1 Max
Qwen3.8 27B27B · Intelligence 33.7	4-bit	15.2 GB	2–4	12–21
Qwen3.6 35B A3B35B · 3B active · Intelligence 18.2	4-bit	19.7 GB	32–56	106–188
Muse Glimmer 30B29.6B · Intelligence 17.5	4-bit	16.7 GB	3–6	11–19
Gemma 4 31B30.7B · Intelligence 14.7	4-bit	17.3 GB	3–6	10–18
Nemotron 3.5 Lightning 30B A3B30B · 3B active · Intelligence 12.9	4-bit	16.9 GB	32–56	106–188
Granite 4.2 30B30B · Intelligence 12.7	4-bit	16.9 GB	3–6	11–19
MiniCPM5 2B2.52B · Intelligence 12.5	8-bit	2.7 GB	20–36	67–119
Qwen3.5 9B9B · Intelligence 11.2	8-bit	9.6 GB	11–19	19–33
Granite 4.2 8B8B · Intelligence 11.1	8-bit	8.5 GB	12–21	21–38
Granite 4.2 3B3B · Intelligence 9.1	8-bit	3.2 GB	17–30	56–100
gpt-oss-20b21B · 3.6B active · Intelligence 9	native 4.25-bit	13.8 GB	28–51	95–168
Nemotron 3 Nano 30B A3B30B · 3.5B active · Intelligence 8.9	4-bit	16.9 GB	27–48	90–160
Gemma 4 E4B8B · Intelligence 8.9	8-bit	8.5 GB	11–20	38–67
Ministral 3 8B8.9B · Intelligence 5.5	8-bit	9.5 GB	11–19	19–34
Nemotron 3 Nano 4B3.97B · Intelligence 4.8	8-bit	4.2 GB	13–23	43–76
Qwen3 Coder 30B A3B Instruct30.5B · 3.3B active	4-bit	17.2 GB	28–51	95–168
GLM-4.7 Flash30B · 3B active	4-bit	16.9 GB	32–56	106–188
Qwen3.6 27B27B	4-bit	15.2 GB	2–4	12–21
Gemma 4 26B A4B25.2B · 3.8B active	6-bit	20.5 GB	17–31	58–103
Devstral Small 2 (24B)24B	4-bit	13.5 GB	3–5	13–24
Magistral Small 1.2 (2509)24B	4-bit	13.5 GB	3–5	13–24
Mistral Small 3.2 (24B)24B	4-bit	13.5 GB	3–5	13–24
LFM2.5 24B A2B24B · 2B active	6-bit	19.5 GB	34–60	113–200
Phi-4 Reasoning Vision 15B15B	8-bit	15.9 GB	3–6	11–20
Phi-4 (14B)14B	8-bit	14.9 GB	4–6	12–21
Gemma 4 12B Unified12B	8-bit	12.7 GB	4–8	14–25
Qwen3 8B8.2B	8-bit	8.7 GB	12–21	21–37
Llama 3.1 8B Instruct8B	8-bit	8.5 GB	12–21	21–38
Gemma 4 E2B5.1B	8-bit	5.4 GB	23–40	75–133
Qwen3.5 4B4B	8-bit	4.3 GB	13–22	42–74
Phi-4 mini instruct3.8B	8-bit	4 GB	14–24	45–80
Llama 3.2 3B Instruct3.21B	8-bit	3.4 GB	16–28	53–94
SmolLM3 3B3B	8-bit	3.2 GB	17–30	56–100
LFM2.5 2.6B2.69B	8-bit	2.9 GB	19–33	62–110
Qwen3.5 2B2B	8-bit	2.1 GB	26–46	86–152
Llama 3.2 1B Instruct1.23B	8-bit	1.3 GB	42–74	138–246
Qwen3.5 0.8B0.8B	8-bit	0.9 GB	60–107	200–356
Qwen3 0.6B0.6B	8-bit	0.6 GB	90–160	300–533

Runs, but slowly 1

Fits, but under 10 tokens a second: fine for batch work, slow for chat.

Model	Version	Memory	M4	M1 Max
Llama 3.3 70B Instruct70B	2-bit lower quality	21.9 GB	2–4	8–15

Only just fits 2

Needs almost all memory: raise the GPU memory limit and keep the context short.

Model	Version	Memory	M4	M1 Max
Nemotron 3.5 Bolt 74B A9B74B · 9B active · Intelligence 21.4	2-bit lower quality	23.1 GB	19–34	64–114
Qwen3 Coder Next80B · 3B active · Intelligence 9.2	2-bit lower quality	25 GB	60–107	200–356

Questions

What is the best LLM for a Mac with 32 GB?: Of the models with a published Intelligence Index, Qwen3.8 27B scores highest among those that run well in 32 GB (4-bit). The biggest that runs well is Qwen3.6 35B A3B. For code: Qwen3 Coder 30B A3B Instruct.
Is 32 GB enough for local AI on a Mac?: 38 of the 73 open models here run well in it at 4 bits or better. macOS lets the GPU use about 24 GB of it by default; the model and its context have to fit in that.
Can a 32 GB Mac run gpt-oss-20b?: Yes, native 4.25-bit, at about 28–51 tokens a second on an M4 and 95–168 on an M1 Max.

Other sizes: 8 GB · 16 GB · 18 GB · 24 GB · 32 GB · 36 GB · 48 GB · 64 GB · 96 GB · 128 GB · 192 GB · 256 GB · 512 GB

Run them with MLX Harness (free, open source) or Apple’s mlx-lm in Terminal: step by step.

How it’s worked out

Memory: the model’s parameters at 8, 6, 4, 3 or 2 bits per weight plus MLX’s group scales (half a bit per weight), or its native format, with room for the context on top. By default macOS lets the GPU use about 75% of unified memory. Speed: memory bandwidth divided by the bytes read per token is the ceiling for one conversation; in llama.cpp’s public Apple Silicon benchmarks real runs reach about 45% to 80% of it, which is the range shown. These are estimates, not measurements: measure your own Mac on MLX Bench. Model figures from each official model card, chip figures from Apple’s Tech Specs. Checked 2026-10-19. Same data as mlx.app/compare.

The best LLMs for a 32 GB Mac.