Arao AI
Back to search results

Qwen3-Embedding-0.6B · IQ1_S

Start a discussion
AMD Ryzen AI Max+ 395128GBllama.cpp v0.36L2 Cross-framework verified

This page aggregates 1 real-world runs of Qwen3-Embedding-0.6B (IQ1_S) on AMD Ryzen AI Max+ 395 with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.

285.55 tok/s

Decode

Decode speed

1,529.78 tok/s

Prefill

Prefill speed

1.12 s

TTFT

Time to first token

2.62 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 · IQ1_S (current)Decode 285.55 · Prefill 1529.78 ·
  2. Ollama · Q5_K_S Decode 220.14 · Prefill 748.53 ·
  3. vLLM · IQ3_XS Decode 324.5 · Prefill 1240.61 ·
  4. Strata · IQ2_M Decode 293.11 · Prefill 1444.58 ·
  5. Strata · Q3_K_L Decode 322.37 · Prefill 1428.17 ·

Core figures

Decode (avg)
285.55 tok/s
Prefill (avg)
1,529.78 tok/s
TTFT (avg)
1.12 s
VRAM (avg)
2.62 GB
MTP acceptance rate
—
TTFB
2.28 GB
Power draw
90.9 W

Configuration

Member-level fields are taken from the most recent run

Model
Qwen3-Embedding-0.6B
Quantization
IQ1_S
Framework
llama.cpp
Version
v0.36
Context length
4096 tokens
Batch size
512
GPU layers
99
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)
2.62 GB
OS
macOS 15.0
Driver
570.86
CUDA
N/A
Power draw
90.9 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

    v0.36 · macOS 15.0 · CUDA N/A · 4096 ctx

    285.55 tok/s

    Decode

    1,529.78 tok/s

    Prefill

    1.12 s

    TTFT

    2.62 GB

    VRAM

    —

    MTP

    90.9 W

    Power draw

    L2 batch seed 2026-10-16: llama.cpp on hw_id=24 running model_id=62 at IQ1_S