Arao AI
Back to search results

Qwen3.5-0.8B · IQ2_XS

Start a discussion
Apple M3 Max36GBllama.cpp v0.32L2 Cross-framework verified

This page aggregates 1 real-world runs of Qwen3.5-0.8B (IQ2_XS) on Apple M3 Max with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.

369.47 tok/s

Decode

Decode speed

2,026.25 tok/s

Prefill

Prefill speed

1.42 s

TTFT

Time to first token

2.91 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 (current)Decode 369.47 · Prefill 2026.25 ·
  2. llama.cpp · IQ3_XS Decode 313.86 · Prefill 1064.12 ·
  3. Ollama · Q3_K_M Decode 286.62 · Prefill 1285.71 ·
  4. Strata · Q8_0 Decode 237.48 · Prefill 1069.27 ·

Core figures

Decode (avg)
369.47 tok/s
Prefill (avg)
2,026.25 tok/s
TTFT (avg)
1.42 s
VRAM (avg)
2.91 GB
MTP acceptance rate
—
TTFB
2.96 GB
Power draw
37.7 W

Configuration

Member-level fields are taken from the most recent run

Model
Qwen3.5-0.8B
Quantization
IQ2_XS
Framework
llama.cpp
Version
v0.32
Context length
8192 tokens
Batch size
256
GPU layers
99
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)
2.91 GB
OS
macOS 15.0
Driver
Apple M3 Max
CUDA
N/A
Power draw
37.7 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

    v0.32 · macOS 15.0 · CUDA N/A · 8192 ctx

    369.47 tok/s

    Decode

    2,026.25 tok/s

    Prefill

    1.42 s

    TTFT

    2.91 GB

    VRAM

    —

    MTP

    37.7 W

    Power draw

    L2 batch seed 2026-10-10: llama.cpp on hw_id=4 running model_id=51 at IQ2_XS