What is the best LLM for a Mac with 128 GB?

Of the models with a published Intelligence Index, Qwen3.8 27B scores highest among those that run well in 128 GB (8-bit). The biggest that runs well is Qwen3.5 122B A10B. For code: Qwen3 Coder Next.

Is 128 GB enough for local AI on a Mac?

47 of the 73 open models here run well in it at 4 bits or better. macOS lets the GPU use about 96 GB of it by default; the model and its context have to fit in that.

Can a 128 GB Mac run gpt-oss-20b?

Yes, native 4.25-bit, at about 95–168 tokens a second on an M3 Max and 189–337 on an M1 Ultra.

Best local LLMs for a 128 GB Mac (M1 to M5): what runs, how fast

Macs with 128 GB

5 chips from M1 to M5 come with 128 GB. Open one for its own list and speeds.

Runs well 51

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

Model	Version	Memory	M3 Max	M1 Ultra
Qwen3.8 27B27B · Intelligence 33.7	8-bit	28.7 GB	12–21	13–22
Nemotron 3.5 Bolt 74B A9B74B · 9B active · Intelligence 21.4	8-bit	78.6 GB	19–33	38–67
Qwen3.6 35B A3B35B · 3B active · Intelligence 18.2	8-bit	37.2 GB	56–100	113–200
Muse Glimmer 30B29.6B · Intelligence 17.5	8-bit	31.5 GB	11–19	11–20
Qwen3.5 122B A10B122B · 10B active · Intelligence 15.6	4-bit	68.6 GB	32–57	64–114
Gemma 4 31B30.7B · Intelligence 14.7	8-bit	32.6 GB	10–18	11–20
Nemotron 3.5 Lightning 30B A3B30B · 3B active · Intelligence 12.9	8-bit	31.9 GB	56–100	113–200
Granite 4.2 30B30B · Intelligence 12.7	8-bit	31.9 GB	11–19	11–20
MiniCPM5 2B2.52B · Intelligence 12.5	8-bit	2.7 GB	67–119	133–237
gpt-oss-120b117B · 5.1B active · Intelligence 11.6	native 4.25-bit	65.3 GB	67–119	133–237
Mistral Small 4 (119B)119B · 6.5B active · Intelligence 11.3	4-bit	66.9 GB	49–86	97–173
Qwen3.5 9B9B · Intelligence 11.2	8-bit	9.6 GB	19–33	38–67
Granite 4.2 8B8B · Intelligence 11.1	8-bit	8.5 GB	21–38	42–75
Qwen3 Coder Next80B · 3B active · Intelligence 9.2	8-bit	85 GB	56–100	113–200
Granite 4.2 3B3B · Intelligence 9.1	8-bit	3.2 GB	56–100	113–200
gpt-oss-20b21B · 3.6B active · Intelligence 9	native 4.25-bit	13.8 GB	95–168	189–337
Nemotron 3 Nano 30B A3B30B · 3.5B active · Intelligence 8.9	8-bit	31.9 GB	49–86	97–173
Gemma 4 E4B8B · Intelligence 8.9	8-bit	8.5 GB	38–67	75–133
Llama 4 Scout 17B-16E Instruct109B · 17B active · Intelligence 6.5	6-bit	88.6 GB	13–23	26–46
Ministral 3 8B8.9B · Intelligence 5.5	8-bit	9.5 GB	19–34	38–67
Nemotron 3 Nano 4B3.97B · Intelligence 4.8	8-bit	4.2 GB	43–76	86–152
Ornith 1.5 122B122B · 10B active	4-bit	68.6 GB	32–57	64–114
Nemotron 3 Super 120B A12B120B · 12B active	4-bit	67.5 GB	26–47	53–94
GLM-4.5 Air106B · 12B active	6-bit	86.1 GB	18–33	37–65
Qwen3 Coder 30B A3B Instruct30.5B · 3.3B active	8-bit	32.4 GB	51–91	103–183
GLM-4.7 Flash30B · 3B active	8-bit	31.9 GB	56–100	113–200
Qwen3.6 27B27B	8-bit	28.7 GB	12–21	13–22
Gemma 4 26B A4B25.2B · 3.8B active	8-bit	26.8 GB	45–80	90–160
Devstral Small 2 (24B)24B	8-bit	25.5 GB	13–24	14–25
Magistral Small 1.2 (2509)24B	8-bit	25.5 GB	13–24	14–25
Mistral Small 3.2 (24B)24B	8-bit	25.5 GB	13–24	14–25
LFM2.5 24B A2B24B · 2B active	8-bit	25.5 GB	86–152	171–305
Phi-4 Reasoning Vision 15B15B	8-bit	15.9 GB	11–20	23–40
Phi-4 (14B)14B	8-bit	14.9 GB	12–21	24–43
Gemma 4 12B Unified12B	8-bit	12.7 GB	14–25	28–50
Qwen3 8B8.2B	8-bit	8.7 GB	21–37	41–74
Llama 3.1 8B Instruct8B	8-bit	8.5 GB	21–38	42–75
Gemma 4 E2B5.1B	8-bit	5.4 GB	75–133	150–267
Qwen3.5 4B4B	8-bit	4.3 GB	42–74	84–149
Phi-4 mini instruct3.8B	8-bit	4 GB	45–80	90–160
Llama 3.2 3B Instruct3.21B	8-bit	3.4 GB	53–94	106–188
SmolLM3 3B3B	8-bit	3.2 GB	56–100	113–200
LFM2.5 2.6B2.69B	8-bit	2.9 GB	62–110	124–221
Qwen3.5 2B2B	8-bit	2.1 GB	86–152	171–305
Llama 3.2 1B Instruct1.23B	8-bit	1.3 GB	138–246	277–492
Qwen3.5 0.8B0.8B	8-bit	0.9 GB	200–356	400–711
Qwen3 0.6B0.6B	8-bit	0.6 GB	300–533	600–1067
Qwen3.8 Flash Next180B · 6B active · Intelligence 39.8	3-bit lower quality	78.8 GB	69–123	138–246
Inkling Small276B · 12B active · Intelligence 25.7	2-bit lower quality	86.3 GB	47–84	95–168
DeepSeek V4 Flash 0731284B · 13B active	2-bit lower quality	88.8 GB	44–78	88–156
MiniMax M2.7230B · 10B active	2-bit lower quality	71.9 GB	58–103	116–206

Runs, but slowly 3

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

Model	Version	Memory	M3 Max	M1 Ultra
Mistral Medium 3.5 (128B)128B · Intelligence 14.2	4-bit	72 GB	3–4	5–9
Devstral 2 (123B)123B	4-bit	69.2 GB	3–5	5–9
Llama 3.3 70B Instruct70B	8-bit	74.4 GB	2–4	5–9

Only just fits 3

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

Model	Version	Memory	M3 Max	M1 Ultra
GLM-5.3 Flash320B · 18B active · Intelligence 41.8	2-bit lower quality	100 GB	32–57	64–114
MiMo V2.6 Flash309B · 15B active · Intelligence 37.9	2-bit lower quality	96.6 GB	38–68	77–136
Hy3295B · 21B active · Intelligence 25.3	2-bit lower quality	92.2 GB	27–48	55–97

Too big 16

Needs more memory than 128 GB, even at 2 bits.

Questions

What is the best LLM for a Mac with 128 GB?: Of the models with a published Intelligence Index, Qwen3.8 27B scores highest among those that run well in 128 GB (8-bit). The biggest that runs well is Qwen3.5 122B A10B. For code: Qwen3 Coder Next.
Is 128 GB enough for local AI on a Mac?: 47 of the 73 open models here run well in it at 4 bits or better. macOS lets the GPU use about 96 GB of it by default; the model and its context have to fit in that.
Can a 128 GB Mac run gpt-oss-20b?: Yes, native 4.25-bit, at about 95–168 tokens a second on an M3 Max and 189–337 on an M1 Ultra.

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.