Arao AI
Back to search results

Qwen3.6-35B-A3B · IQ1_S

Start a discussion
Apple M3 Max36GBvLLM v0.36L2 Cross-framework verified

This page aggregates 1 real-world runs of Qwen3.6-35B-A3B (IQ1_S) on Apple M3 Max with vLLM, contributed by 1 independent source platforms; metrics are averages of published measurements.

102.56 tok/s

Decode

Decode speed

411.25 tok/s

Prefill

Prefill speed

1.72 s

TTFT

Time to first token

32.97 GB

VRAM

VRAM usage

L2 Cross-framework verified

1

Measured runs

1

Independent sources

Other

Source platforms

Today

Last verified

Performance

  1. llama.cpp · IQ2_XS Decode 89.41 · Prefill 406.89 ·
  2. vLLM · IQ1_S (current)Decode 102.56 · Prefill 411.25 ·

Core figures

Decode (avg)
102.56 tok/s
Prefill (avg)
411.25 tok/s
TTFT (avg)
1.72 s
VRAM (avg)
32.97 GB
MTP acceptance rate
—
TTFB
0.82 GB
Power draw
86.4 W

Configuration

Member-level fields are taken from the most recent run

Model
Qwen3.6-35B-A3B
Quantization
IQ1_S
Framework
vLLM
Version
v0.36
Context length
16384 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 M3 Max
Nominal VRAM
36 GB
Measured VRAM (avg)
32.97 GB
OS
macOS 15.0
Driver
Apple M3 Max
CUDA
N/A
Power draw
86.4 W

Sources & evidence

1 measured records in total, each traceable to its original source

  1. L2 Cross-framework verifiedOtherOriginal link Verified on 2026-10-10

    v0.36 · macOS 15.0 · CUDA N/A · 16384 ctx

    102.56 tok/s

    Decode

    411.25 tok/s

    Prefill

    1.72 s

    TTFT

    32.97 GB

    VRAM

    —

    MTP

    86.4 W

    Power draw

    L2 batch seed 2026-10-10: vLLM on hw_id=4 running model_id=47 at IQ1_S