AP Apple Prosumer Unified memory

Apple M2 Ultra 192GB

Runs downloadable LLMs up to ~298B at Q4, ~121.4-146.4 tok/s on an 8B model.

Other configs: 64GB
Apple M2 Ultra 192GB
192GB
Unified memory
800
GB/s bandwidth
298B
Max model (Q4)
$6,600
MSRP

What you can run

At 8k context, FP16 KV. Full-speed tok/s holds up to "max ctx"; beyond it the context spills to SSD swap and drops to the degraded figure.

Model sizeFP16Q8Q4
3B 90.7-109.3 t/s
≤2522k ctx · then ~1.1-1.4 (slow)
181.3-218.7 t/s
≤2568k ctx · then ~2.3-2.7 (slow)
323.8-390.5 t/s
≤2588k ctx · then ~4-4.9 (slow)
7-8B 34-41 t/s
≤1185k ctx · then ~0.4-0.5 (slow)
68-82 t/s
≤1246k ctx · then ~0.9-1 (slow)
121.4-146.4 t/s
≤1273k ctx · then ~1.5-1.8 (slow)
13B 20.9-25.2 t/s
≤887k ctx · then ~0.3-0.3 (slow)
41.8-50.5 t/s
≤966k ctx · then ~0.5-0.6 (slow)
74.7-90.1 t/s
≤1001k ctx · then ~0.9-1.1 (slow)
32B 8.5-10.3 t/s
≤409k ctx · then ~0.1-0.1 (slow)
17-20.5 t/s
≤531k ctx · then ~0.2-0.3 (slow)
30.4-36.6 t/s
≤585k ctx · then ~0.4-0.5 (slow)
70B 3.9-4.7 t/s
≤96k ctx · then ~0-0.1 (slow)
7.8-9.4 t/s
≤309k ctx · then ~0.1-0.1 (slow)
13.9-16.7 t/s
≤403k ctx · then ~0.2-0.2 (slow)
120B won't run 4.5-5.5 t/s
≤157k ctx · then ~0.1-0.1 (slow)
8.1-9.8 t/s
≤318k ctx · then ~0.1-0.1 (slow)
400B won't run won't run won't run

Estimate: tok/s ≈ bandwidth ÷ (params × bytes/param) × efficiency (Apple). Cyan = fits ~172.8 GB usable. Beyond "ctx", KV spills to SSD swap (rough degraded speed shown).

Run popular models on the Apple M2 Ultra 192GB

Videos about the Apple M2 Ultra 192GB

Apple fans, start typing your angry comments now… M2 Ultra Mac Studio Review

Mac Studio M2 Ultra Review - It DESTROYS my MacBook Pro! 🔥

More hardware like the Apple M2 Ultra 192GB

About the Apple M2 Ultra 192GB

Apple M2 Ultra 192GB is a unified-memory system from Apple on Apple M2, 2023. For local genAI the numbers that matter are its 192 GB (how big a model + context fits) and 800 GB/s bandwidth (generation speed).

Frequently asked questions

What size LLM can the Apple M2 Ultra 192GB run?

With ~172.8 GB usable it runs up to ~298B at Q4 (~84B at FP16) at 8k context. Bigger spills into SSD swap and slows sharply.

How many tokens/sec does the Apple M2 Ultra 192GB generate?

Generation is bandwidth-bound. At 800 GB/s an 8B model at Q4 runs ~121.4-146.4 tok/s (estimate).

Why does long context slow it down?

The KV cache for the context must also fit fast memory; once weights+KV exceed 172.8 GB it spills to SSD swap and throughput collapses. Each row shows the max usable context.