Arao AI
Back to search results

Qwen3.8-27B · FP16

Start a discussion
AMD Ryzen AI Max+ 395128GBvLLM v0.45L2 Cross-framework verified

This page aggregates 1 real-world runs of Qwen3.8-27B (FP16) on AMD Ryzen AI Max+ 395 with vLLM, contributed by 1 independent source platforms; metrics are averages of published measurements.

9.75 tok/s

Decode

Decode speed

49.87 tok/s

Prefill

Prefill speed

2.04 s

TTFT

Time to first token

107 GB

VRAM

VRAM usage

L2 Cross-framework verified

1

Measured runs

1

Independent sources

Other

Source platforms

17 days ago

Last verified

Performance

  1. vLLM · FP16 (current)Decode 9.75 · Prefill 49.87 ·
  2. vLLM · Q8_0 Decode 13.23 · Prefill 57.54 ·
  3. Ollama · Q3_K_L Decode 19.34 · Prefill 100.78 ·

Core figures

Decode (avg)
9.75 tok/s
Prefill (avg)
49.87 tok/s
TTFT (avg)
2.04 s
VRAM (avg)
107 GB
MTP acceptance rate
—
TTFB
0.08 GB
Power draw
64.3 W

Configuration

Member-level fields are taken from the most recent run

Model
Qwen3.8-27B
Quantization
FP16
Framework
vLLM
Version
v0.45
Context length
8192 tokens
Batch size
16
GPU layers
999
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)
107 GB
OS
Windows 11 23H2
Driver
amdgpu-pro 24.20
CUDA
12.6
Power draw
64.3 W

Get started

Original model resources

Base model resources may not match this quantization. Use the verified deployment weights above when available.

Sources & evidence

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

  1. L2 Cross-framework verifiedOtherOriginal link Verified on 2026-10-19

    v0.45 · Windows 11 23H2 · CUDA 12.6 · 8192 ctx

    9.75 tok/s

    Decode

    49.87 tok/s

    Prefill

    2.04 s

    TTFT

    107 GB

    VRAM

    —

    MTP

    64.3 W

    Power draw

    L2 batch seed 2026-10-19: vLLM on hw_id=24 running model_id=5 at FP16