Arao AI
Back to search results

Qwen3-4B · Q6_K

Start a discussion
AMD Radeon RX 9070 XT16GBllama.cpp v0.33L2 Cross-framework verified

This page aggregates 1 real-world runs of Qwen3-4B (Q6_K) on AMD Radeon RX 9070 XT with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.

94.45 tok/s

Decode

Decode speed

349.72 tok/s

Prefill

Prefill speed

1.82 s

TTFT

Time to first token

10.5 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 · Q2_K Decode 162 · Prefill 560.55 ·
  2. llama.cpp · Q6_K (current)Decode 94.45 · Prefill 349.72 ·
  3. Ollama · IQ3_S Decode 124.75 · Prefill 481.56 ·

Core figures

Decode (avg)
94.45 tok/s
Prefill (avg)
349.72 tok/s
TTFT (avg)
1.82 s
VRAM (avg)
10.5 GB
MTP acceptance rate
—
TTFB
0.76 GB
Power draw
309.2 W

Configuration

Member-level fields are taken from the most recent run

Model
Qwen3-4B
Quantization
Q6_K
Framework
llama.cpp
Version
v0.33
Context length
16384 tokens
Batch size
512
GPU layers
20
Flash Attention
On

Hardware

Nominal and measured figures are shown side by side; whether it runs is the reader's call

GPU
AMD Radeon RX 9070 XT
Nominal VRAM
16 GB
Measured VRAM (avg)
10.5 GB
OS
Windows 11 23H2
Driver
amdgpu-pro 24.20
CUDA
N/A
Power draw
309.2 W

Sources & evidence

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

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

    v0.33 · Windows 11 23H2 · CUDA N/A · 16384 ctx

    94.45 tok/s

    Decode

    349.72 tok/s

    Prefill

    1.82 s

    TTFT

    10.5 GB

    VRAM

    —

    MTP

    309.2 W

    Power draw

    L2 batch seed 2026-10-13: llama.cpp on hw_id=16 running model_id=54 at Q6_K