Arao AI
Back to search results

Llama-3.2-3B-Instruct · Q8_0

Start a discussion
AMD Ryzen AI Max+ 395128GBllama.cpp v0.24L2 Cross-framework verified

This page aggregates 1 real-world runs of Llama-3.2-3B-Instruct (Q8_0) on AMD Ryzen AI Max+ 395 with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.

61.59 tok/s

Decode

Decode speed

327.72 tok/s

Prefill

Prefill speed

1.3 s

TTFT

Time to first token

9.35 GB

VRAM

VRAM usage

L2 Cross-framework verified

1

Measured runs

1

Independent sources

Other

Source platforms

16 days ago

Last verified

Performance

  1. llama.cpp · Q8_0 (current)Decode 61.59 · Prefill 327.72 ·
  2. vLLM · Q4_K_M Decode 89.28 · Prefill 327.44 ·

Core figures

Decode (avg)
61.59 tok/s
Prefill (avg)
327.72 tok/s
TTFT (avg)
1.3 s
VRAM (avg)
9.35 GB
MTP acceptance rate
—
TTFB
0.49 GB
Power draw
60.8 W

Configuration

Member-level fields are taken from the most recent run

Model
Llama-3.2-3B-Instruct
Quantization
Q8_0
Framework
llama.cpp
Version
v0.24
Context length
16384 tokens
Batch size
512
GPU layers
28
Flash Attention
On

Hardware

Nominal and measured figures are shown side by side; whether it runs is the reader's call

GPU
AMD Ryzen AI Max+ 395
Nominal VRAM
128 GB
Measured VRAM (avg)
9.35 GB
OS
Ubuntu 24.04
Driver
amdgpu 6.11
CUDA
12.4
Power draw
60.8 W

Sources & evidence

1 measured records in total, each traceable to its original source

  1. L2 Cross-framework verifiedOtherOriginal link Verified on 2026-10-19

    v0.24 · Ubuntu 24.04 · CUDA 12.4 · 16384 ctx

    61.59 tok/s

    Decode

    327.72 tok/s

    Prefill

    1.3 s

    TTFT

    9.35 GB

    VRAM

    —

    MTP

    60.8 W

    Power draw

    L2 batch seed 2026-10-19: llama.cpp on hw_id=24 running model_id=70 at Q8_0