Arao AI
Back to search results

Llama-3.1-8B-Instruct · Q4_K_S

Start a discussion
Apple M2 UltraOllama v0.41L2 Cross-framework verified

This page aggregates 1 real-world runs of Llama-3.1-8B-Instruct (Q4_K_S) on Apple M2 Ultra with Ollama, contributed by 1 independent source platforms; metrics are averages of published measurements.

103.18 tok/s

Decode

Decode speed

542.74 tok/s

Prefill

Prefill speed

2.1 s

TTFT

Time to first token

15.2 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 · Q3_K_L Decode 129.48 · Prefill 476.49 ·
  2. llama.cpp · Q4_K_S Decode 123.15 · Prefill 540.31 ·
  3. Ollama · Q4_K_S (current)Decode 103.18 · Prefill 542.74 ·

Core figures

Decode (avg)
103.18 tok/s
Prefill (avg)
542.74 tok/s
TTFT (avg)
2.1 s
VRAM (avg)
15.2 GB
MTP acceptance rate
—
TTFB
0.83 GB
Power draw
128.4 W

Configuration

Member-level fields are taken from the most recent run

Model
Llama-3.1-8B-Instruct
Quantization
Q4_K_S
Framework
Ollama
Version
v0.41
Context length
32768 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)
15.2 GB
OS
macOS 26.1
Power draw
128.4 W

Sources & evidence

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

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

    v0.41 · macOS 26.1 · 32768 ctx

    103.18 tok/s

    Decode

    542.74 tok/s

    Prefill

    2.1 s

    TTFT

    15.2 GB

    VRAM

    —

    MTP

    128.4 W

    Power draw

    L2 batch seed 2026-10-19: Ollama on hw_id=103 running model_id=68 at Q4_K_S