Arao AI
Back to search results

Qwen2.5-Omni-3B · IQ4_XS

Start a discussion
Apple M4 Pro64GBvLLM v0.40L2 Cross-framework verified

This page aggregates 1 real-world runs of Qwen2.5-Omni-3B (IQ4_XS) on Apple M4 Pro with vLLM, contributed by 1 independent source platforms; metrics are averages of published measurements.

172.49 tok/s

Decode

Decode speed

661.58 tok/s

Prefill

Prefill speed

0.74 s

TTFT

Time to first token

6.76 GB

VRAM

VRAM usage

L2 Cross-framework verified

1

Measured runs

1

Independent sources

Other

Source platforms

5 days ago

Last verified

Performance

  1. vLLM · IQ4_XS (current)Decode 172.49 · Prefill 661.58 ·
  2. vLLM · Q3_K_S Decode 184.21 · Prefill 1008.8 ·
  3. Ollama · Q4_K_S Decode 154.32 · Prefill 813.63 ·
  4. Ollama · Q5_K_M Decode 114.09 · Prefill 438.42 ·

Core figures

Decode (avg)
172.49 tok/s
Prefill (avg)
661.58 tok/s
TTFT (avg)
0.74 s
VRAM (avg)
6.76 GB
MTP acceptance rate
—
TTFB
1.38 GB
Power draw
95 W

Configuration

Member-level fields are taken from the most recent run

Model
Qwen2.5-Omni-3B
Quantization
IQ4_XS
Framework
vLLM
Version
v0.40
Context length
2048 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
Apple M4 Pro
Nominal VRAM
64 GB
Measured VRAM (avg)
6.76 GB
OS
macOS 15.0
Driver
Apple M4 Pro
CUDA
N/A
Power draw
95 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

    v0.40 · macOS 15.0 · CUDA N/A · 2048 ctx

    172.49 tok/s

    Decode

    661.58 tok/s

    Prefill

    0.74 s

    TTFT

    6.76 GB

    VRAM

    —

    MTP

    95 W

    Power draw

    L2 batch seed 2026-10-13: vLLM on hw_id=25 running model_id=60 at IQ4_XS