Arao AI
Back to search results

Qwen3.5-0.8B · IQ4_XS

Start a discussion
Apple M1 MaxvLLM v0.34L2 Cross-framework verified

This page aggregates 1 real-world runs of Qwen3.5-0.8B (IQ4_XS) on Apple M1 Max with vLLM, contributed by 1 independent source platforms; metrics are averages of published measurements.

222.4 tok/s

Decode

Decode speed

1,142.41 tok/s

Prefill

Prefill speed

1.1 s

TTFT

Time to first token

3.34 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 · IQ4_XS (current)Decode 222.4 · Prefill 1142.41 ·
  2. vLLM · Q3_K_S Decode 238.61 · Prefill 1236.73 ·
  3. Ollama · IQ3_S Decode 214.42 · Prefill 1034.95 ·
  4. Ollama · Q8_0 Decode 124.21 · Prefill 540.32 ·

Core figures

Decode (avg)
222.4 tok/s
Prefill (avg)
1,142.41 tok/s
TTFT (avg)
1.1 s
VRAM (avg)
3.34 GB
MTP acceptance rate
—
TTFB
1.78 GB
Power draw
42.2 W

Configuration

Member-level fields are taken from the most recent run

Model
Qwen3.5-0.8B
Quantization
IQ4_XS
Framework
vLLM
Version
v0.34
Context length
4096 tokens
Batch size
16
GPU layers
999
Flash Attention
Off

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)
3.34 GB
OS
macOS 15.0
Driver
Apple M1 Max
CUDA
N/A
Power draw
42.2 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 · macOS 15.0 · CUDA N/A · 4096 ctx

    222.4 tok/s

    Decode

    1,142.41 tok/s

    Prefill

    1.1 s

    TTFT

    3.34 GB

    VRAM

    —

    MTP

    42.2 W

    Power draw

    L2 batch seed 2026-10-16: vLLM on hw_id=101 running model_id=51 at IQ4_XS