192GB VRAM

Mac M3 Ultra 192G — what LLMs can it run?

74 of 79 indexed models fit comfortably in 192GB at 4K context, each at the highest-quality quant that still leaves headroom.

70B+ · 10

Mac M3 Ultra 192G — what LLMs can it run? — 70B+
Model	Quant	Est. VRAM	Headroom	tok/s
Qwen3 235B-A22B Instruct235B-A22B	Q4_K_M	149.68 GB	+42.3 GB	10
DBRX Instruct132B	Q4_K_M	84.31 GB	+107.7 GB	15
GPT-OSS 120B117B MoE	MXFP4	65.15 GB	+126.9 GB	22
Llama 4 Scout 17B (16E)109B MoE	Q4_K_M	69.88 GB	+122.1 GB	22
GLM-4.5-Air106B MoE	Q4_K_M	67.94 GB	+124.1 GB	18
Llama 3.2 90B Vision Instruct90B	Q4_K_M	57.5 GB	+134.5 GB	22
Qwen2.5 72B Instruct72B	Q5_K_M	55.31 GB	+136.7 GB	24
Llama 3.1 70B Instruct70B	Q5_K_M	53.75 GB	+138.3 GB	32
Llama 3.3 70B Instruct70B	Q5_K_M	53.75 GB	+138.3 GB	32
DeepSeek-R1-Distill-Llama-70B70B	Q4_K_M	46.1 GB	+145.9 GB	36

32B · 18

Mac M3 Ultra 192G — what LLMs can it run? — 32B
Model	Quant	Est. VRAM	Headroom	tok/s
Mixtral 8x7B Instruct47B MoE	Q4_K_M	30.13 GB	+161.9 GB	48
Seed-OSS 36B Instruct36B	Q5_K_M	27.81 GB	+164.2 GB	37
Command R 35B35B	Q4_K_M	22.86 GB	+169.1 GB	42
Yi 1.5 34B Chat34B	Q4_K_M	22.82 GB	+169.2 GB	40
Qwen3 32B Instruct32B	Q4_K_M	21.85 GB	+170.2 GB	42
Qwen2.5 32B Instruct32B	Q4_K_M	21.69 GB	+170.3 GB	44
Qwen2.5-Coder 32B Instruct32B	Q4_K_M	21.69 GB	+170.3 GB	44
DeepSeek-R1-Distill-Qwen-32B32B	Q4_K_M	21.69 GB	+170.3 GB	42
Qwen3 30B-A3B Instruct30B-A3B	Q5_K_M	23.04 GB	+169 GB	82
Qwen3-Coder 30B-A3B Instruct30B-A3B	Q5_K_M	23.04 GB	+169 GB	80
Qwen3-VL 30B-A3B Instruct30B-A3B	Q8_0	34.28 GB	+157.7 GB	78
Gemma 3 27B IT27B	Q4_K_M	19.49 GB	+172.5 GB	48
Gemma 2 27B Instruct27B	Q5_K_M	21.76 GB	+170.2 GB	42
Mistral Small 24B Instruct24B	EXL2 4.65bpw	15.26 GB	+176.7 GB	88
Devstral Small 1.1 24B24B	Q6_K	20.91 GB	+171.1 GB	48
Magistral Small 1.2 24B24B	Q6_K	20.91 GB	+171.1 GB	47
Codestral 22B22B	Q4_K_M	15.03 GB	+177 GB	58
GPT-OSS 20B21B MoE	MXFP4	11.81 GB	+180.2 GB	195

14B · 17

Mac M3 Ultra 192G — what LLMs can it run? — 14B
Model	Quant	Est. VRAM	Headroom	tok/s
InternLM2 20B Chat20B	Q5_K_M	15.59 GB	+176.4 GB	68
DeepSeek-Coder-V2-Lite Instruct16B	Q8_0	18.36 GB	+173.6 GB	118
DeepSeek-V2-Lite Chat16B	Q4_K_M	10.87 GB	+181.1 GB	142
StarCoder2 15B15B	Q4_K_M	10.19 GB	+181.8 GB	92
Qwen3 14B Instruct14B	Q5_K_M	11.65 GB	+180.4 GB	78
Qwen2.5 14B Instruct14B	Q5_K_M	11.73 GB	+180.3 GB	86
DeepSeek-R1-Distill-Qwen-14B14B	EXL2 4.65bpw	9.75 GB	+182.3 GB	128
Phi-4 14B14B	Q5_K_M	11.77 GB	+180.2 GB	78
Phi-3 Medium 14B Instruct14B	Q6_K	12.86 GB	+179.1 GB	88
Mistral Nemo 12B Instruct12B	Q6_K	11.14 GB	+180.9 GB	95
Gemma 3 12B IT12B	Q5_K_M	10.6 GB	+181.4 GB	92
Stable LM 2 12B Chat12B	Q4_K_M	8.35 GB	+183.7 GB	108
Jamba 1.5 Mini12B	Q4_K_M	8.15 GB	+183.9 GB	95
Llama 3.2 11B Vision Instruct11B	Q8_0	12.9 GB	+179.1 GB	72
Solar 10.7B Instruct11B	Q4_K_M	7.6 GB	+184.4 GB	125
Falcon 3 10B Instruct10B	Q4_K_M	7.28 GB	+184.7 GB	118
Gemma 2 9B Instruct9B	Q8_0	11.7 GB	+180.3 GB	108

7B · 19

Mac M3 Ultra 192G — what LLMs can it run? — 7B
Model	Quant	Est. VRAM	Headroom	tok/s
GLM-4-9B-Chat9B	Q8_0	10.16 GB	+181.8 GB	105
Qwen3-VL 8B Instruct8B	Q8_0	10.39 GB	+181.6 GB	108
Qwen2-VL 7B Instruct7B	Q4_K_M	5.49 GB	+186.5 GB	72
Granite 3.1 8B Instruct8B	Q4_K_M	5.88 GB	+186.1 GB	142
Qwen3 8B Instruct8B	Q6_K	7.64 GB	+184.4 GB	122
Llama 3.1 8B Instruct8B	Q8_0	9.47 GB	+182.5 GB	118
Nous Hermes 3 Llama 3.1 8B8B	EXL2 4.65bpw	5.43 GB	+186.6 GB	232
Aya 23 8B8B	Q4_K_M	5.64 GB	+186.4 GB	145
OpenChat 3.6 8B8B	EXL2 4.65bpw	5.43 GB	+186.6 GB	228
DeepSeek-R1-Distill-Llama-8B8B	Q5_K_M	6.51 GB	+185.5 GB	128
InternLM2 7B Chat7B	Q4_K_M	5.45 GB	+186.6 GB	148
Qwen2.5 7B Instruct7B	Q6_K	6.77 GB	+185.2 GB	132
Qwen2.5-Coder 7B Instruct7B	EXL2 4.65bpw	4.87 GB	+187.1 GB	248
WizardLM-2 7B7B	Q4_K_M	5.07 GB	+186.9 GB	152
DeepSeek-R1-Distill-Qwen-7B7B	EXL2 4.65bpw	4.87 GB	+187.1 GB	210
OLMo 2 7B Instruct7B	Q8_0	10.3 GB	+181.7 GB	125
Mistral 7B Instruct v0.37B	Q6_K	6.75 GB	+185.3 GB	135
Zephyr 7B Beta7B	Q6_K	6.75 GB	+185.3 GB	132
Gemma 3 4B IT4B	Q8_0	5.36 GB	+186.6 GB	145

≤3B · 10

Mac M3 Ultra 192G — what LLMs can it run? — ≤3B
Model	Quant	Est. VRAM	Headroom	tok/s
Qwen3 4B Instruct4B	Q6_K	4.05 GB	+188 GB	145
Phi-4 Mini Instruct3.8B	Q8_0	4.68 GB	+187.3 GB	262
Phi-3.5 Mini Instruct3.8B	Q8_0	5.89 GB	+186.1 GB	255
Llama 3.2 3B Instruct3B	Q8_0	4.05 GB	+188 GB	285
Qwen2.5 3B Instruct3B	Q8_0	3.67 GB	+188.3 GB	290
Gemma 2 2B Instruct2B	Q8_0	3.34 GB	+188.7 GB	320
Qwen3 1.7B Instruct1.7B	Q8_0	2.39 GB	+189.6 GB	240
Qwen2.5 1.5B Instruct1.5B	Q8_0	1.83 GB	+190.2 GB	410
Llama 3.2 1B Instruct1B	Q8_0	1.51 GB	+190.5 GB	450
Qwen2.5 0.5B Instruct0.5B	Q8_0	0.6 GB	+191.4 GB	540

How this list is built

Each row is the lowest-perplexity-loss quant of that model whose estimated total — weights plus KV cache at 4K context plus activation buffer — uses at most 88% of the card. That is the calculator's "green" threshold, so every row here has real headroom rather than only just fitting. Raise the context length and the list shortens; the calculator lets you check any combination directly.

Open this card in the VRAM calculator Browse all 79 models