Arao AI
Back to search results

Qwen3-8B · Q8_0

Start a discussion
Apple M2 UltraOllama v0.50L2 Cross-framework verified

This page aggregates 1 real-world runs of Qwen3-8B (Q8_0) on Apple M2 Ultra with Ollama, contributed by 1 independent source platforms; metrics are averages of published measurements.

67.83 tok/s

Decode

Decode speed

284.74 tok/s

Prefill

Prefill speed

0.35 s

TTFT

Time to first token

21.42 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 · Q4_K_S Decode 108.67 · Prefill 436.07 ·
  2. llama.cpp · Q5_K_S Decode 87.67 · Prefill 476.02 ·
  3. Ollama · Q5_K_M Decode 79.98 · Prefill 403.29 ·
  4. Ollama · Q8_0 (current)Decode 67.83 · Prefill 284.74 ·

Core figures

Decode (avg)
67.83 tok/s
Prefill (avg)
284.74 tok/s
TTFT (avg)
0.35 s
VRAM (avg)
21.42 GB
MTP acceptance rate
—
TTFB
0.54 GB
Power draw
67.2 W

Configuration

Member-level fields are taken from the most recent run

Model
Qwen3-8B
Quantization
Q8_0
Framework
Ollama
Version
v0.50
Context length
16384 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 M2 Ultra
Measured VRAM (avg)
21.42 GB
OS
macOS 15.0
Driver
Apple M2 Ultra
CUDA
N/A
Power draw
67.2 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

    v0.50 · macOS 15.0 · CUDA N/A · 16384 ctx

    67.83 tok/s

    Decode

    284.74 tok/s

    Prefill

    0.35 s

    TTFT

    21.42 GB

    VRAM

    —

    MTP

    67.2 W

    Power draw

    L2 batch seed 2026-10-16: Ollama on hw_id=103 running model_id=53 at Q8_0