Arao AI
Back to search results

Qwen3.5-4B · Q5_K_M

Start a discussion
AMD Strix Point (Radeon 890M)llama.cpp v0.37L2 Cross-framework verified

This page aggregates 1 real-world runs of Qwen3.5-4B (Q5_K_M) on AMD Strix Point (Radeon 890M) with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.

58.41 tok/s

Decode

Decode speed

300.39 tok/s

Prefill

Prefill speed

2.5 s

TTFT

Time to first token

9.91 GB

VRAM

VRAM usage

L2 Cross-framework verified

1

Measured runs

1

Independent sources

Other

Source platforms

2 days ago

Last verified

Performance

  1. llama.cpp · Q5_K_M (current)Decode 58.41 · Prefill 300.39 ·
  2. Ollama · Q3_K_L Decode 78.29 · Prefill 430.35 ·
  3. Ollama · Q8_0 Decode 43.81 · Prefill 179.92 ·

Core figures

Decode (avg)
58.41 tok/s
Prefill (avg)
300.39 tok/s
TTFT (avg)
2.5 s
VRAM (avg)
9.91 GB
MTP acceptance rate
—
TTFB
0.47 GB
Power draw
67.4 W

Configuration

Member-level fields are taken from the most recent run

Model
Qwen3.5-4B
Quantization
Q5_K_M
Framework
llama.cpp
Version
v0.37
Context length
2048 tokens
Batch size
512
GPU layers
28
Flash Attention
Off

Hardware

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

GPU
AMD Strix Point (Radeon 890M)
Measured VRAM (avg)
9.91 GB
OS
Windows 11 23H2
Driver
amdgpu 6.11
CUDA
N/A
Power draw
67.4 W

Sources & evidence

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

  1. L2 Cross-framework verifiedOtherOriginal link Verified on 2026-10-16

    v0.37 · Windows 11 23H2 · CUDA N/A · 2048 ctx

    58.41 tok/s

    Decode

    300.39 tok/s

    Prefill

    2.5 s

    TTFT

    9.91 GB

    VRAM

    —

    MTP

    67.4 W

    Power draw

    L2 batch seed 2026-10-16: llama.cpp on hw_id=89 running model_id=48 at Q5_K_M