Arao AI
Back to search results

Qwen3.5-0.8B · IQ2_XS

Start a discussion
AMD Ryzen AI Max+ 395128GBllama.cpp v0.20L2 Cross-framework verified

This page aggregates 1 real-world runs of Qwen3.5-0.8B (IQ2_XS) on AMD Ryzen AI Max+ 395 with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.

354.37 tok/s

Decode

Decode speed

1,453.65 tok/s

Prefill

Prefill speed

0.12 s

TTFT

Time to first token

2.91 GB

VRAM

VRAM usage

L2 Cross-framework verified

1

Measured runs

1

Independent sources

Other

Source platforms

5 days ago

Last verified

Performance

  1. llama.cpp · IQ4_XS Decode 273.83 · Prefill 1312.19 ·
  2. llama.cpp · Q5_K_M Decode 197.9 · Prefill 695.65 ·
  3. Ollama · Q4_K_M Decode 220.93 · Prefill 741.5 ·
  4. Ollama · Q8_0 Decode 163.22 · Prefill 720.03 ·
  5. llama.cpp · IQ1_S Decode 380.54 · Prefill 1677.44 ·
  6. llama.cpp · IQ2_XS (current)Decode 354.37 · Prefill 1453.65 ·

Core figures

Decode (avg)
354.37 tok/s
Prefill (avg)
1,453.65 tok/s
TTFT (avg)
0.12 s
VRAM (avg)
2.91 GB
MTP acceptance rate
—
TTFB
2.83 GB
Power draw
81.5 W

Configuration

Member-level fields are taken from the most recent run

Model
Qwen3.5-0.8B
Quantization
IQ2_XS
Framework
llama.cpp
Version
v0.20
Context length
16384 tokens
Batch size
256
GPU layers
33
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)
2.91 GB
OS
Windows 11 23H2
Driver
AMD Adrenalin 25.10.1
CUDA
N/A
Power draw
81.5 W

Sources & evidence

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

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

    v0.20 · Windows 11 23H2 · CUDA N/A · 16384 ctx

    354.37 tok/s

    Decode

    1,453.65 tok/s

    Prefill

    0.12 s

    TTFT

    2.91 GB

    VRAM

    —

    MTP

    81.5 W

    Power draw

    L2 batch seed 2026-10-13: llama.cpp on hw_id=24 running model_id=51 at IQ2_XS