Arao AI
Back to search results

Qwen3.5-9B · IQ3_XS

Start a discussion
AMD Ryzen AI Max+ 395128GBllama.cpp v0.45L2 Cross-framework verified

This page aggregates 1 real-world runs of Qwen3.5-9B (IQ3_XS) on AMD Ryzen AI Max+ 395 with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.

118.29 tok/s

Decode

Decode speed

392.28 tok/s

Prefill

Prefill speed

1.58 s

TTFT

Time to first token

13.82 GB

VRAM

VRAM usage

L2 Cross-framework verified

1

Measured runs

1

Independent sources

Other

Source platforms

5 days ago

Last verified

Performance

  1. vLLM · IQ3_XS Decode 115.94 · Prefill 451.93 ·
  2. vLLM · Q5_K_M Decode 83.86 · Prefill 287.36 ·
  3. Ollama · Q4_K_S Decode 85.51 · Prefill 324.93 ·
  4. llama.cpp · IQ3_XS (current)Decode 118.29 · Prefill 392.28 ·

Core figures

Decode (avg)
118.29 tok/s
Prefill (avg)
392.28 tok/s
TTFT (avg)
1.58 s
VRAM (avg)
13.82 GB
MTP acceptance rate
—
TTFB
0.95 GB
Power draw
118.3 W

Configuration

Member-level fields are taken from the most recent run

Model
Qwen3.5-9B
Quantization
IQ3_XS
Framework
llama.cpp
Version
v0.45
Context length
4096 tokens
Batch size
512
GPU layers
99
Flash Attention
Off

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)
13.82 GB
OS
Windows 11 23H2
Driver
AMD Adrenalin 25.10.1
CUDA
N/A
Power draw
118.3 W

Sources & evidence

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

  1. L2 Cross-framework verifiedOtherOriginal link Verified on 2026-10-13

    v0.45 · Windows 11 23H2 · CUDA N/A · 4096 ctx

    118.29 tok/s

    Decode

    392.28 tok/s

    Prefill

    1.58 s

    TTFT

    13.82 GB

    VRAM

    —

    MTP

    118.3 W

    Power draw

    L2 batch seed 2026-10-13: llama.cpp on hw_id=24 running model_id=49 at IQ3_XS