Arao AI
Back to search results

gemma-4-12B-it · Q8_0

Start a discussion
Apple M5 Pro64GBllama.cpp v0.24L2 Cross-framework verified

This page aggregates 1 real-world runs of gemma-4-12B-it (Q8_0) on Apple M5 Pro with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.

66.87 tok/s

Decode

Decode speed

301.78 tok/s

Prefill

Prefill speed

1.96 s

TTFT

Time to first token

31.08 GB

VRAM

VRAM usage

L2 Cross-framework verified

1

Measured runs

1

Independent sources

Other

Source platforms

5 days ago

Last verified

Performance

  1. llama.cpp · Q4_K_M Decode 100.37 · Prefill 374.04 ·
  2. llama.cpp · Q8_0 (current)Decode 66.87 · Prefill 301.78 ·
  3. Ollama · Q3_K_M Decode 110.36 · Prefill 569.49 ·
  4. Ollama · Q3_K_S Decode 111.33 · Prefill 497.48 ·

Core figures

Decode (avg)
66.87 tok/s
Prefill (avg)
301.78 tok/s
TTFT (avg)
1.96 s
VRAM (avg)
31.08 GB
MTP acceptance rate
—
TTFB
0.53 GB
Power draw
144.2 W

Configuration

Member-level fields are taken from the most recent run

Model
gemma-4-12B-it
Quantization
Q8_0
Framework
llama.cpp
Version
v0.24
Context length
4096 tokens
Batch size
512
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 M5 Pro
Nominal VRAM
64 GB
Measured VRAM (avg)
31.08 GB
OS
macOS 15.0
Driver
Apple M5 Pro
CUDA
N/A
Power draw
144.2 W

Sources & evidence

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

  1. L2 Cross-framework verifiedOtherOriginal link Verified on 2026-10-13

    v0.24 · macOS 15.0 · CUDA N/A · 4096 ctx

    66.87 tok/s

    Decode

    301.78 tok/s

    Prefill

    1.96 s

    TTFT

    31.08 GB

    VRAM

    —

    MTP

    144.2 W

    Power draw

    L2 batch seed 2026-10-13: llama.cpp on hw_id=27 running model_id=65 at Q8_0