Arao AI
Back to search results

Qwen3.5-9B · IQ3_XS

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

This page aggregates 1 real-world runs of Qwen3.5-9B (IQ3_XS) on Apple M1 Pro with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.

75.72 tok/s

Decode

Decode speed

277.89 tok/s

Prefill

Prefill speed

0.55 s

TTFT

Time to first token

13.82 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 · IQ3_XS (current)Decode 75.72 · Prefill 277.89 ·
  2. vLLM · IQ2_M Decode 84.15 · Prefill 281.3 ·
  3. vLLM · Q3_K_S Decode 79.35 · Prefill 395.79 ·

Core figures

Decode (avg)
75.72 tok/s
Prefill (avg)
277.89 tok/s
TTFT (avg)
0.55 s
VRAM (avg)
13.82 GB
MTP acceptance rate
—
TTFB
0.61 GB
Power draw
45.3 W

Configuration

Member-level fields are taken from the most recent run

Model
Qwen3.5-9B
Quantization
IQ3_XS
Framework
llama.cpp
Version
v0.34
Context length
32768 tokens
Batch size
256
GPU layers
28
Flash Attention
Off

Hardware

Nominal and measured figures are shown side by side; whether it runs is the reader's call

GPU
Apple M1 Pro
Measured VRAM (avg)
13.82 GB
OS
macOS 15.0
Driver
Apple M1 Pro
CUDA
N/A
Power draw
45.3 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

    v0.34 · macOS 15.0 · CUDA N/A · 32768 ctx

    75.72 tok/s

    Decode

    277.89 tok/s

    Prefill

    0.55 s

    TTFT

    13.82 GB

    VRAM

    —

    MTP

    45.3 W

    Power draw

    L2 batch seed 2026-10-16: llama.cpp on hw_id=100 running model_id=49 at IQ3_XS