Arao AI
Back to search results

Qwen3-4B-Thinking-2507 · Q8_0

Start a discussion
AMD Ryzen AI Max+ 395128GBllama.cpp v0.51L2 Cross-framework verified

This page aggregates 1 real-world runs of Qwen3-4B-Thinking-2507 (Q8_0) on AMD Ryzen AI Max+ 395 with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.

78.29 tok/s

Decode

Decode speed

263.52 tok/s

Prefill

Prefill speed

0.96 s

TTFT

Time to first token

11.76 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 Decode 107.36 · Prefill 478 ·
  2. llama.cpp · Q8_0 (current)Decode 78.29 · Prefill 263.52 ·
  3. Ollama · Q3_K_S Decode 145.01 · Prefill 581.64 ·
  4. llama.cpp · IQ2_M Decode 178.19 · Prefill 583.74 ·
  5. Strata · Q3_K_L Decode 190.53 · Prefill 851.53 ·

Core figures

Decode (avg)
78.29 tok/s
Prefill (avg)
263.52 tok/s
TTFT (avg)
0.96 s
VRAM (avg)
11.76 GB
MTP acceptance rate
—
TTFB
0.63 GB
Power draw
118.4 W

Configuration

Member-level fields are taken from the most recent run

Model
Qwen3-4B-Thinking-2507
Quantization
Q8_0
Framework
llama.cpp
Version
v0.51
Context length
16384 tokens
Batch size
256
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)
11.76 GB
OS
macOS 15.0
Driver
560.94
CUDA
N/A
Power draw
118.4 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

    v0.51 · macOS 15.0 · CUDA N/A · 16384 ctx

    78.29 tok/s

    Decode

    263.52 tok/s

    Prefill

    0.96 s

    TTFT

    11.76 GB

    VRAM

    —

    MTP

    118.4 W

    Power draw

    L2 batch seed 2026-10-16: llama.cpp on hw_id=24 running model_id=57 at Q8_0