Arao AI
Back to search results

Qwen3.6-35B-A3B · Q8_0

Start a discussion
Apple M2 Ultrallama.cpp v0.54L2 Cross-framework verified

This page aggregates 1 real-world runs of Qwen3.6-35B-A3B (Q8_0) on Apple M2 Ultra with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.

114.98 tok/s

Decode

Decode speed

558.64 tok/s

Prefill

Prefill speed

1.11 s

TTFT

Time to first token

76 GB

VRAM

VRAM usage

L2 Cross-framework verified

1

Measured runs

1

Independent sources

Other

Source platforms

16 days ago

Last verified

Performance

  1. llama.cpp · IQ1_S Decode 250.48 · Prefill 981.93 ·
  2. llama.cpp · Q8_0 (current)Decode 114.98 · Prefill 558.64 ·
  3. Ollama · Q2_K Decode 217.13 · Prefill 942.96 ·

Core figures

Decode (avg)
114.98 tok/s
Prefill (avg)
558.64 tok/s
TTFT (avg)
1.11 s
VRAM (avg)
76 GB
MTP acceptance rate
—
TTFB
0.92 GB
Power draw
120.8 W

Configuration

Member-level fields are taken from the most recent run

Model
Qwen3.6-35B-A3B
Quantization
Q8_0
Framework
llama.cpp
Version
v0.54
Context length
2048 tokens
Batch size
256
GPU layers
33
Flash Attention
Off

Hardware

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

GPU
Apple M2 Ultra
Measured VRAM (avg)
76 GB
OS
macOS 26.1
Power draw
120.8 W

Sources & evidence

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

  1. L2 Cross-framework verifiedOtherOriginal link Verified on 2026-10-19

    v0.54 · macOS 26.1 · 2048 ctx

    114.98 tok/s

    Decode

    558.64 tok/s

    Prefill

    1.11 s

    TTFT

    76 GB

    VRAM

    —

    MTP

    120.8 W

    Power draw

    L2 batch seed 2026-10-19: llama.cpp on hw_id=103 running model_id=47 at Q8_0