What is the best LLM for a Mac with 96 GB?

Of the models with a published Intelligence Index, Qwen3.8 27B scores highest among those that run well in 96 GB (8-bit). The biggest that runs well is gpt-oss-120b. For code: Qwen3 Coder Next.

Is 96 GB enough for local AI on a Mac?

44 of the 73 open models here run well in it at 4 bits or better. macOS lets the GPU use about 72 GB of it by default; the model and its context have to fit in that.

Can a 96 GB Mac run gpt-oss-20b?

Yes, native 4.25-bit, at about 71–126 tokens a second on an M3 Max and 284–505 on an M5 Ultra.

Best local LLMs for a 96 GB Mac (M2 to M5): what runs, how fast

Macs with 96 GB

4 chips from M2 to M5 come with 96 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 M5 Ultra.

Model	Version	Memory	M3 Max	M5 Ultra
Qwen3.8 27B27B · Intelligence 33.7	8-bit	28.7 GB	5–8	19–33
Nemotron 3.5 Bolt 74B A9B74B · 9B active · Intelligence 21.4	6-bit	60.1 GB	18–33	74–132
Qwen3.6 35B A3B35B · 3B active · Intelligence 18.2	8-bit	37.2 GB	42–75	169–300
Muse Glimmer 30B29.6B · Intelligence 17.5	8-bit	31.5 GB	4–8	17–30
Gemma 4 31B30.7B · Intelligence 14.7	8-bit	32.6 GB	4–7	17–29
Nemotron 3.5 Lightning 30B A3B30B · 3B active · Intelligence 12.9	8-bit	31.9 GB	42–75	169–300
Granite 4.2 30B30B · Intelligence 12.7	8-bit	31.9 GB	4–8	17–30
MiniCPM5 2B2.52B · Intelligence 12.5	8-bit	2.7 GB	50–89	200–356
gpt-oss-120b117B · 5.1B active · Intelligence 11.6	native 4.25-bit	65.3 GB	50–89	200–356
Qwen3.5 9B9B · Intelligence 11.2	8-bit	9.6 GB	14–25	56–100
Granite 4.2 8B8B · Intelligence 11.1	8-bit	8.5 GB	16–28	64–113
Qwen3 Coder Next80B · 3B active · Intelligence 9.2	6-bit	65 GB	56–100	225–400
Granite 4.2 3B3B · Intelligence 9.1	8-bit	3.2 GB	42–75	169–300
gpt-oss-20b21B · 3.6B active · Intelligence 9	native 4.25-bit	13.8 GB	71–126	284–505
Nemotron 3 Nano 30B A3B30B · 3.5B active · Intelligence 8.9	8-bit	31.9 GB	36–65	146–259
Gemma 4 E4B8B · Intelligence 8.9	8-bit	8.5 GB	28–50	113–200
Llama 4 Scout 17B-16E Instruct109B · 17B active · Intelligence 6.5	4-bit	61.3 GB	14–25	56–100
Ministral 3 8B8.9B · Intelligence 5.5	8-bit	9.5 GB	14–25	57–101
Nemotron 3 Nano 4B3.97B · Intelligence 4.8	8-bit	4.2 GB	32–57	129–229
GLM-4.5 Air106B · 12B active	4-bit	59.6 GB	20–35	79–141
Llama 3.3 70B Instruct70B	4-bit	39.4 GB	2–4	14–24
Qwen3 Coder 30B A3B Instruct30.5B · 3.3B active	8-bit	32.4 GB	39–69	154–274
GLM-4.7 Flash30B · 3B active	8-bit	31.9 GB	42–75	169–300
Qwen3.6 27B27B	8-bit	28.7 GB	5–8	19–33
Gemma 4 26B A4B25.2B · 3.8B active	8-bit	26.8 GB	34–60	135–240
Devstral Small 2 (24B)24B	8-bit	25.5 GB	10–18	21–38
Magistral Small 1.2 (2509)24B	8-bit	25.5 GB	10–18	21–38
Mistral Small 3.2 (24B)24B	8-bit	25.5 GB	10–18	21–38
LFM2.5 24B A2B24B · 2B active	8-bit	25.5 GB	64–114	257–457
Phi-4 Reasoning Vision 15B15B	8-bit	15.9 GB	11–20	34–60
Phi-4 (14B)14B	8-bit	14.9 GB	12–21	36–64
Gemma 4 12B Unified12B	8-bit	12.7 GB	11–19	43–76
Qwen3 8B8.2B	8-bit	8.7 GB	16–28	62–110
Llama 3.1 8B Instruct8B	8-bit	8.5 GB	16–28	64–113
Gemma 4 E2B5.1B	8-bit	5.4 GB	56–100	225–400
Qwen3.5 4B4B	8-bit	4.3 GB	31–56	126–223
Phi-4 mini instruct3.8B	8-bit	4 GB	34–60	135–240
Llama 3.2 3B Instruct3.21B	8-bit	3.4 GB	40–71	159–282
SmolLM3 3B3B	8-bit	3.2 GB	42–75	169–300
LFM2.5 2.6B2.69B	8-bit	2.9 GB	47–83	186–331
Qwen3.5 2B2B	8-bit	2.1 GB	64–114	257–457
Llama 3.2 1B Instruct1.23B	8-bit	1.3 GB	104–185	415–738
Qwen3.5 0.8B0.8B	8-bit	0.9 GB	150–267	600–1067
Qwen3 0.6B0.6B	8-bit	0.6 GB	225–400	900–1600
Qwen3.8 Flash Next180B · 6B active · Intelligence 39.8	2-bit lower quality	56.3 GB	71–126	284–505
Qwen3.5 122B A10B122B · 10B active · Intelligence 15.6	3-bit lower quality	53.4 GB	31–55	123–218
Mistral Medium 3.5 (128B)128B · Intelligence 14.2	3-bit lower quality	56 GB	2–4	10–17
Mistral Small 4 (119B)119B · 6.5B active · Intelligence 11.3	3-bit lower quality	52.1 GB	48–86	193–343
Devstral 2 (123B)123B	3-bit lower quality	53.8 GB	3–4	10–18
Ornith 1.5 122B122B · 10B active	3-bit lower quality	53.4 GB	31–55	123–218
Nemotron 3 Super 120B A12B120B · 12B active	3-bit lower quality	52.5 GB	25–45	102–181

Only just fits 1

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

Model	Version	Memory	M3 Max	M5 Ultra
MiniMax M2.7230B · 10B active	2-bit lower quality	71.9 GB	44–77	174–310

Questions

What is the best LLM for a Mac with 96 GB?: Of the models with a published Intelligence Index, Qwen3.8 27B scores highest among those that run well in 96 GB (8-bit). The biggest that runs well is gpt-oss-120b. For code: Qwen3 Coder Next.
Is 96 GB enough for local AI on a Mac?: 44 of the 73 open models here run well in it at 4 bits or better. macOS lets the GPU use about 72 GB of it by default; the model and its context have to fit in that.
Can a 96 GB Mac run gpt-oss-20b?: Yes, native 4.25-bit, at about 71–126 tokens a second on an M3 Max and 284–505 on an M5 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.

The best LLMs for a 96 GB Mac.