Arao AI
Back to search results

Qwen3-4B-Thinking-2507 · Q3_K_S

Start a discussion
AMD Ryzen AI Max+ 395128GBOllama v0.53L2 Cross-framework verified

This page aggregates 1 real-world runs of Qwen3-4B-Thinking-2507 (Q3_K_S) on AMD Ryzen AI Max+ 395 with Ollama, contributed by 1 independent source platforms; metrics are averages of published measurements.

145.01 tok/s

Decode

Decode speed

581.64 tok/s

Prefill

Prefill speed

2 s

TTFT

Time to first token

7.56 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 Decode 78.29 · Prefill 263.52 ·
  3. Ollama · Q3_K_S (current)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)
145.01 tok/s
Prefill (avg)
581.64 tok/s
TTFT (avg)
2 s
VRAM (avg)
7.56 GB
MTP acceptance rate
—
TTFB
1.16 GB
Power draw
98.2 W

Configuration

Member-level fields are taken from the most recent run

Model
Qwen3-4B-Thinking-2507
Quantization
Q3_K_S
Framework
Ollama
Version
v0.53
Context length
4096 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)
7.56 GB
OS
macOS 15.0
Driver
570.86
CUDA
N/A
Power draw
98.2 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

    v0.53 · macOS 15.0 · CUDA N/A · 4096 ctx

    145.01 tok/s

    Decode

    581.64 tok/s

    Prefill

    2 s

    TTFT

    7.56 GB

    VRAM

    —

    MTP

    98.2 W

    Power draw

    L2 batch seed 2026-10-16: Ollama on hw_id=24 running model_id=57 at Q3_K_S