Arao AI
Back to search results

Llama-3.1-8B-Instruct · IQ3_XS

Start a discussion
AMD Ryzen AI Max+ 395128GBllama.cpp v0.26L2 Cross-framework verified

This page aggregates 1 real-world runs of Llama-3.1-8B-Instruct (IQ3_XS) on AMD Ryzen AI Max+ 395 with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.

45.88 tok/s

Decode

Decode speed

214.9 tok/s

Prefill

Prefill speed

0.24 s

TTFT

Time to first token

12.52 GB

VRAM

VRAM usage

L2 Cross-framework verified

1

Measured runs

1

Independent sources

Other

Source platforms

17 days ago

Last verified

Performance

  1. llama.cpp · IQ3_XS (current)Decode 45.88 · Prefill 214.9 ·
  2. vLLM · Q2_K Decode 51.17 · Prefill 219.96 ·

Core figures

Decode (avg)
45.88 tok/s
Prefill (avg)
214.9 tok/s
TTFT (avg)
0.24 s
VRAM (avg)
12.52 GB
MTP acceptance rate
—
TTFB
0.37 GB
Power draw
46.5 W

Configuration

Member-level fields are taken from the most recent run

Model
Llama-3.1-8B-Instruct
Quantization
IQ3_XS
Framework
llama.cpp
Version
v0.26
Context length
32768 tokens
Batch size
512
GPU layers
99
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)
12.52 GB
OS
Ubuntu 22.04
Driver
amdgpu-pro 24.20
CUDA
12.6
Power draw
46.5 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

    v0.26 · Ubuntu 22.04 · CUDA 12.6 · 32768 ctx

    45.88 tok/s

    Decode

    214.9 tok/s

    Prefill

    0.24 s

    TTFT

    12.52 GB

    VRAM

    —

    MTP

    46.5 W

    Power draw

    L2 batch seed 2026-10-19: llama.cpp on hw_id=24 running model_id=68 at IQ3_XS