Arao AI
Back to search results

Llama-3.1-8B-Instruct · Q3_K_S

Start a discussion
Apple M5 Ultra512GBllama.cpp v0.29L2 Cross-framework verified

This page aggregates 1 real-world runs of Llama-3.1-8B-Instruct (Q3_K_S) on Apple M5 Ultra with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.

134.78 tok/s

Decode

Decode speed

545.26 tok/s

Prefill

Prefill speed

1.99 s

TTFT

Time to first token

13.02 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_S (current)Decode 134.78 · Prefill 545.26 ·
  2. llama.cpp · Q5_K_S Decode 104.75 · Prefill 395.8 ·
  3. Ollama · Q8_0 Decode 65.71 · Prefill 314.08 ·

Core figures

Decode (avg)
134.78 tok/s
Prefill (avg)
545.26 tok/s
TTFT (avg)
1.99 s
VRAM (avg)
13.02 GB
MTP acceptance rate
—
TTFB
1.08 GB
Power draw
96.2 W

Configuration

Member-level fields are taken from the most recent run

Model
Llama-3.1-8B-Instruct
Quantization
Q3_K_S
Framework
llama.cpp
Version
v0.29
Context length
2048 tokens
Batch size
256
GPU layers
28
Flash Attention
On

Hardware

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

GPU
Apple M5 Ultra
Nominal VRAM
512 GB
Measured VRAM (avg)
13.02 GB
OS
macOS 26.1
Power draw
96.2 W

Sources & evidence

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

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

    v0.29 · macOS 26.1 · 2048 ctx

    134.78 tok/s

    Decode

    545.26 tok/s

    Prefill

    1.99 s

    TTFT

    13.02 GB

    VRAM

    —

    MTP

    96.2 W

    Power draw

    L2 batch seed 2026-10-19: llama.cpp on hw_id=29 running model_id=68 at Q3_K_S