Arao AI
Back to search results

Qwen2.5-Omni-3B · Q4_K_S

Start a discussion
AMD Ryzen AI Max+ 395128GBllama.cpp v0.55L2 Cross-framework verified

This page aggregates 1 real-world runs of Qwen2.5-Omni-3B (Q4_K_S) on AMD Ryzen AI Max+ 395 with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.

153.58 tok/s

Decode

Decode speed

827.06 tok/s

Prefill

Prefill speed

0.9 s

TTFT

Time to first token

7.01 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_S Decode 201.15 · Prefill 1047.45 ·
  2. vLLM · Q2_K Decode 204.49 · Prefill 918.23 ·
  3. Strata · Q4_K_M Decode 218.64 · Prefill 754.88 ·
  4. Strata · Q8_0 Decode 147.88 · Prefill 774.09 ·
  5. llama.cpp · Q4_K_S (current)Decode 153.58 · Prefill 827.06 ·
  6. llama.cpp · Q8_0 Decode 98.01 · Prefill 478.15 ·
  7. Ollama · Q3_K_S Decode 173.92 · Prefill 636.41 ·
  8. Ollama · Q4_K_S Decode 150.76 · Prefill 704.53 ·

Core figures

Decode (avg)
153.58 tok/s
Prefill (avg)
827.06 tok/s
TTFT (avg)
0.9 s
VRAM (avg)
7.01 GB
MTP acceptance rate
—
TTFB
1.23 GB
Power draw
143 W

Configuration

Member-level fields are taken from the most recent run

Model
Qwen2.5-Omni-3B
Quantization
Q4_K_S
Framework
llama.cpp
Version
v0.55
Context length
32768 tokens
Batch size
512
GPU layers
20
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)
7.01 GB
OS
Windows 11 23H2
Driver
AMD Adrenalin 25.10.1
CUDA
N/A
Power draw
143 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

    v0.55 · Windows 11 23H2 · CUDA N/A · 32768 ctx

    153.58 tok/s

    Decode

    827.06 tok/s

    Prefill

    0.9 s

    TTFT

    7.01 GB

    VRAM

    —

    MTP

    143 W

    Power draw

    L2 batch seed 2026-10-13: llama.cpp on hw_id=24 running model_id=60 at Q4_K_S