Arao AI
Back to search results

Qwen3-Coder-Next · Q4_K_S

Start a discussion
AMD Ryzen AI Max+ 395128GBllama.cpp v0.20L2 Cross-framework verified

This page aggregates 1 real-world runs of Qwen3-Coder-Next (Q4_K_S) on AMD Ryzen AI Max+ 395 with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.

90.68 tok/s

Decode

Decode speed

349.79 tok/s

Prefill

Prefill speed

0.68 s

TTFT

Time to first token

1.05 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 · Q4_K_S (current)Decode 90.68 · Prefill 349.79 ·
  2. Ollama · IQ2_M Decode 108.9 · Prefill 355.25 ·
  3. Ollama · Q3_K_S Decode 98.59 · Prefill 347.51 ·

Core figures

Decode (avg)
90.68 tok/s
Prefill (avg)
349.79 tok/s
TTFT (avg)
0.68 s
VRAM (avg)
1.05 GB
MTP acceptance rate
—
TTFB
0.73 GB
Power draw
114.3 W

Configuration

Member-level fields are taken from the most recent run

Model
Qwen3-Coder-Next
Quantization
Q4_K_S
Framework
llama.cpp
Version
v0.20
Context length
2048 tokens
Batch size
512
GPU layers
28
Flash Attention
On

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)
1.05 GB
OS
Windows 11 23H2
Driver
amdgpu 6.10
CUDA
N/A
Power draw
114.3 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

    v0.20 · Windows 11 23H2 · CUDA N/A · 2048 ctx

    90.68 tok/s

    Decode

    349.79 tok/s

    Prefill

    0.68 s

    TTFT

    1.05 GB

    VRAM

    —

    MTP

    114.3 W

    Power draw

    L2 batch seed 2026-10-16: llama.cpp on hw_id=24 running model_id=58 at Q4_K_S