Arao AI
Back to search results

Qwen2.5-0.5B-Instruct · FP16

Start a discussion
AMD Strix Halo (Radeon 8060S)llama.cpp v0.47L2 Cross-framework verified

This page aggregates 1 real-world runs of Qwen2.5-0.5B-Instruct (FP16) on AMD Strix Halo (Radeon 8060S) with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.

124.48 tok/s

Decode

Decode speed

656.36 tok/s

Prefill

Prefill speed

0.83 s

TTFT

Time to first token

4.04 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 · FP16 (current)Decode 124.48 · Prefill 656.36 ·
  2. llama.cpp · Q6_K Decode 171.39 · Prefill 894.49 ·
  3. vLLM · Q5_K_S Decode 228.54 · Prefill 950.08 ·

Core figures

Decode (avg)
124.48 tok/s
Prefill (avg)
656.36 tok/s
TTFT (avg)
0.83 s
VRAM (avg)
4.04 GB
MTP acceptance rate
—
TTFB
1 GB
Power draw
79.7 W

Configuration

Member-level fields are taken from the most recent run

Model
Qwen2.5-0.5B-Instruct
Quantization
FP16
Framework
llama.cpp
Version
v0.47
Context length
8192 tokens
Batch size
512
GPU layers
33
Flash Attention
Off

Hardware

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

GPU
AMD Strix Halo (Radeon 8060S)
Measured VRAM (avg)
4.04 GB
OS
Ubuntu 24.04
Driver
amdgpu 6.11
CUDA
N/A
Power draw
79.7 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

    v0.47 · Ubuntu 24.04 · CUDA N/A · 8192 ctx

    124.48 tok/s

    Decode

    656.36 tok/s

    Prefill

    0.83 s

    TTFT

    4.04 GB

    VRAM

    —

    MTP

    79.7 W

    Power draw

    L2 batch seed 2026-10-16: llama.cpp on hw_id=87 running model_id=59 at FP16