AM AMD Prosumer Unified memory

AMD Ryzen AI Max+ 395

Runs downloadable LLMs up to ~195B at Q4, ~36-44 tok/s on an 8B model.

AM
128GB · 256 GB/s
128GB
Unified memory
256
GB/s bandwidth
195B
Max model (Q4)
$1,999
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 26.9-32.9 t/s
≤1643k ctx · then ~1.1-1.3 (slow)
53.8-65.7 t/s
≤1689k ctx · then ~2.1-2.6 (slow)
96-117.3 t/s
≤1709k ctx · then ~3.8-4.6 (slow)
7-8B 10.1-12.3 t/s
≤745k ctx · then ~0.4-0.5 (slow)
20.2-24.6 t/s
≤806k ctx · then ~0.8-1 (slow)
36-44 t/s
≤833k ctx · then ~1.4-1.7 (slow)
13B 6.2-7.6 t/s
≤535k ctx · then ~0.2-0.3 (slow)
12.4-15.2 t/s
≤615k ctx · then ~0.5-0.6 (slow)
22.2-27.1 t/s
≤650k ctx · then ~0.9-1.1 (slow)
32B 2.5-3.1 t/s
≤190k ctx · then ~0.1-0.1 (slow)
5-6.2 t/s
≤312k ctx · then ~0.2-0.2 (slow)
9-11 t/s
≤365k ctx · then ~0.4-0.4 (slow)
70B won't run 2.3-2.8 t/s
≤133k ctx · then ~0.1-0.1 (slow)
4.1-5 t/s
≤227k ctx · then ~0.2-0.2 (slow)
120B won't run won't run 2.4-2.9 t/s
≤142k 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 (AMD). Cyan = fits ~115.2 GB usable. Beyond "ctx", KV spills to SSD swap (rough degraded speed shown).

Run popular models on the AMD Ryzen AI Max+ 395

Videos about the AMD Ryzen AI Max+ 395

AMD Strix Halo/Ryzen AI Max+ 395 for AI – An Honest Review

Ryzen AI Max+ Pro 395 Unified Memory? Testing LLM Ollama + LM Studio

More hardware like the AMD Ryzen AI Max+ 395

About the AMD Ryzen AI Max+ 395

AMD Ryzen AI Max+ 395 is a unified-memory system from AMD, 2025. For local genAI the numbers that matter are its 128 GB (how big a model + context fits) and 256 GB/s bandwidth (generation speed).

Frequently asked questions

What size LLM can the AMD Ryzen AI Max+ 395 run?

With ~115.2 GB usable it runs up to ~195B at Q4 (~55.5B at FP16) at 8k context. Bigger spills into SSD swap and slows sharply.

How many tokens/sec does the AMD Ryzen AI Max+ 395 generate?

Generation is bandwidth-bound. At 256 GB/s an 8B model at Q4 runs ~36-44 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 115.2 GB it spills to SSD swap and throughput collapses. Each row shows the max usable context.