Arao AI
Back to search results

Qwen2.5-0.5B-Instruct · Q4_K_S

Start a discussion
Apple M1 Maxllama.cpp v0.41L2 Cross-framework verified

This page aggregates 1 real-world runs of Qwen2.5-0.5B-Instruct (Q4_K_S) on Apple M1 Max with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.

266.38 tok/s

Decode

Decode speed

866.19 tok/s

Prefill

Prefill speed

2.16 s

TTFT

Time to first token

2.92 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 266.38 · Prefill 866.19 ·
  2. Strata · Q3_K_L Decode 395.33 · Prefill 1769.04 ·

Core figures

Decode (avg)
266.38 tok/s
Prefill (avg)
866.19 tok/s
TTFT (avg)
2.16 s
VRAM (avg)
2.92 GB
MTP acceptance rate
—
TTFB
2.13 GB
Power draw
39.3 W

Configuration

Member-level fields are taken from the most recent run

Model
Qwen2.5-0.5B-Instruct
Quantization
Q4_K_S
Framework
llama.cpp
Version
v0.41
Context length
32768 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
Apple M1 Max
Measured VRAM (avg)
2.92 GB
OS
macOS 15.0
Driver
Apple M1 Max
CUDA
N/A
Power draw
39.3 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

    v0.41 · macOS 15.0 · CUDA N/A · 32768 ctx

    266.38 tok/s

    Decode

    866.19 tok/s

    Prefill

    2.16 s

    TTFT

    2.92 GB

    VRAM

    —

    MTP

    39.3 W

    Power draw

    L2 batch seed 2026-10-16: llama.cpp on hw_id=101 running model_id=59 at Q4_K_S