Arao AI
Back to search results

Qwen3-4B-Thinking-2507 · Q3_K_L

Start a discussion
AMD Ryzen AI Max+ 395128GBStrata v0.52L2 Cross-framework verified

This page aggregates 1 real-world runs of Qwen3-4B-Thinking-2507 (Q3_K_L) on AMD Ryzen AI Max+ 395 with Strata, contributed by 1 independent source platforms; metrics are averages of published measurements.

190.53 tok/s

Decode

Decode speed

851.53 tok/s

Prefill

Prefill speed

2.13 s

TTFT

Time to first token

8.15 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 · Q5_K_M Decode 107.36 · Prefill 478 ·
  2. llama.cpp · Q8_0 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 (current)Decode 190.53 · Prefill 851.53 ·

Core figures

Decode (avg)
190.53 tok/s
Prefill (avg)
851.53 tok/s
TTFT (avg)
2.13 s
VRAM (avg)
8.15 GB
MTP acceptance rate
0.56
TTFB
1.52 GB
Power draw
92.4 W

Configuration

Member-level fields are taken from the most recent run

Model
Qwen3-4B-Thinking-2507
Quantization
Q3_K_L
Framework
Strata
Version
v0.52
Context length
2048 tokens
Batch size
1024
GPU layers
8
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)
8.15 GB
OS
Windows 11 23H2
Driver
AMD Adrenalin 25.10.1
CUDA
N/A
Power draw
92.4 W

Sources & evidence

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

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

    v0.52 · Windows 11 23H2 · CUDA N/A · 2048 ctx

    190.53 tok/s

    Decode

    851.53 tok/s

    Prefill

    2.13 s

    TTFT

    8.15 GB

    VRAM

    0.56

    MTP

    92.4 W

    Power draw

    L2 batch seed 2026-10-13: Strata on hw_id=24 running model_id=57 at Q3_K_L