Arao AI
Back to search results

Qwen2.5-0.5B-Instruct · IQ2_XS

Start a discussion
Apple M5 Max128GBvLLM v0.34L2 Cross-framework verified

This page aggregates 1 real-world runs of Qwen2.5-0.5B-Instruct (IQ2_XS) on Apple M5 Max with vLLM, contributed by 1 independent source platforms; metrics are averages of published measurements.

291.82 tok/s

Decode

Decode speed

1,370.02 tok/s

Prefill

Prefill speed

1.18 s

TTFT

Time to first token

2.6 GB

VRAM

VRAM usage

L2 Cross-framework verified

1

Measured runs

1

Independent sources

Other

Source platforms

2 days ago

Last verified

Performance

  1. vLLM · IQ2_XS (current)Decode 291.82 · Prefill 1370.02 ·
  2. Strata · IQ1_S Decode 343.91 · Prefill 1513.7 ·
  3. Strata · Q5_K_S Decode 305.21 · Prefill 1469.7 ·
  4. llama.cpp · IQ2_M Decode 298.63 · Prefill 1141.07 ·
  5. llama.cpp · Q8_0 Decode 366.43 · Prefill 1208.56 ·
  6. vLLM · Q2_K Decode 318.68 · Prefill 1287.78 ·

Core figures

Decode (avg)
291.82 tok/s
Prefill (avg)
1,370.02 tok/s
TTFT (avg)
1.18 s
VRAM (avg)
2.6 GB
MTP acceptance rate
—
TTFB
2.33 GB
Power draw
168.5 W

Configuration

Member-level fields are taken from the most recent run

Model
Qwen2.5-0.5B-Instruct
Quantization
IQ2_XS
Framework
vLLM
Version
v0.34
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 M5 Max
Nominal VRAM
128 GB
Measured VRAM (avg)
2.6 GB
OS
Ubuntu 22.04
Driver
555.85
CUDA
12.8
Power draw
168.5 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

    v0.34 · Ubuntu 22.04 · CUDA 12.8 · 2048 ctx

    291.82 tok/s

    Decode

    1,370.02 tok/s

    Prefill

    1.18 s

    TTFT

    2.6 GB

    VRAM

    —

    MTP

    168.5 W

    Power draw

    L2 batch seed 2026-10-16: vLLM on hw_id=28 running model_id=59 at IQ2_XS