GPU GUIDE · APPLE
Best AI models for the
Apple M3 Ultra (192GB)
The Apple M3 Ultra (192GB) has 192GB of unified memory. Below are the top 30 open-source AI models that fit, ranked by composite benchmark score. Each row shows the best quantization that fits your hardware.
VRAM
192GB
Brand
apple
Models that fit
30
Generation
M3
Top 30 models for the Apple M3 Ultra (192GB)
01
77.9Qwen3.8 2.4T A95B
95.0BBest fit: fp16 · 191.0GBCan I run it? →
02
69.0Qwen3.8 27B
27.0BBest fit: fp16 · 55.0GBCan I run it? →
03
54.0GLM-5
230BBest fit: Q6_K · 190.5GBCan I run it? →
04
52.5Qwen 3.6 Plus
235BBest fit: Q5_K_M · 167.8GBCan I run it? →
05
49.1MiMo V2 Omni
120BBest fit: Q8_0 · 128.5GBCan I run it? →
06
48.3Qwen 3.6 27B
27.8BBest fit: f32 · 112.2GBCan I run it? →
07
45.0Qwen 3 Next 80B A3B
80.0BBest fit: Q8_0 · 86.0GBCan I run it? →
08
44.9Qwen3.5-27B
27.0BBest fit: fp16 · 55.0GBCan I run it? →
09
43.7Qwen 3.6 35B A3B
36.0BBest fit: f32 · 145.0GBCan I run it? →
10
40.6Muse Glimmer 30B
30.0BBest fit: fp16 · 61.0GBCan I run it? →
11
39.9Qwen 3.5 Omni Plus
235BBest fit: Q5_K_M · 167.8GBCan I run it? →
12
37.7Qwen 3.5 35B A3B
36.0BBest fit: f32 · 145.0GBCan I run it? →
13
37.0Gemma 4 31B
31.3BBest fit: f32 · 126.2GBCan I run it? →
14
36.7Devstral 2
24.0BBest fit: f32 · 97.0GBCan I run it? →
15
31.9Gemma 4 26B A4B
26.0BBest fit: fp16 · 53.0GBCan I run it? →
16
31.7Devstral Small 2
7.0BBest fit: f32 · 29.0GBCan I run it? →
17
31.1NVIDIA Nemotron 3 Super 120B A12B BF16
124BBest fit: Q8_0 · 132.3GBCan I run it? →
18
26.7Ministral 3 14B
14.0BBest fit: f32 · 57.0GBCan I run it? →
19
26.0GPT-OSS 120B
120BBest fit: Q8_0 · 128.5GBCan I run it? →
20
25.9gemma 4 12B it
12.0BBest fit: f32 · 49.0GBCan I run it? →
21
25.0Ministral 3 8B
8.0BBest fit: f32 · 33.0GBCan I run it? →
22
25.0Gemma 4 E2B
5.1BBest fit: f32 · 21.4GBCan I run it? →
23
24.6Qwen 3.5 9B
9.7BBest fit: f32 · 39.8GBCan I run it? →
24
24.0Qwen3 VL 235B A22B Instruct
236BBest fit: Q5_K_M · 168.3GBCan I run it? →
25
23.2Qwen 3.5 4B
4.7BBest fit: f32 · 19.8GBCan I run it? →
26
23.2Granite 4.2 8B
8.0BBest fit: fp16 · 17.0GBCan I run it? →
27
21.4Mistral Small 4
119BBest fit: Q8_0 · 127.4GBCan I run it? →
28
21.9Qwen3 235B A22B Thinking 2507
235BBest fit: Q4_K_M · 143.5GBCan I run it? →
29
21.1Devstral 2 2512
125BBest fit: Q8_0 · 133.8GBCan I run it? →
30
20.1Qwen 3 235B A22B Instruct 2507
235BBest fit: Q5_K_M · 167.9GBCan I run it? →
Advertisement
FAQ — running AI on the Apple M3 Ultra (192GB)
How many AI models can the Apple M3 Ultra (192GB) run?
With 192GB of unified memory, the Apple M3 Ultra (192GB) can run 30+ open-source models from our database, including Qwen3.8 2.4T A95B, Qwen3.8 27B, GLM-5.
What's the largest LLM I can run on a Apple M3 Ultra (192GB)?
The biggest model that fits is approximately 236B. Larger models would need to be quantized further or won't fit at all.
Is 192GB of unified memory enough for local AI?
Yes — 192GB comfortably runs most popular open-source models including 30B-class LLMs at Q4_K_M.