Arao AI
Back to search results

gemma-4-12B-it · Q5_K_S

Start a discussion
Apple M2 Ultrallama.cpp v0.42L2 Cross-framework verified

This page aggregates 1 real-world runs of gemma-4-12B-it (Q5_K_S) on Apple M2 Ultra with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.

66.9 tok/s

Decode

Decode speed

326.75 tok/s

Prefill

Prefill speed

1.94 s

TTFT

Time to first token

24.78 GB

VRAM

VRAM usage

L2 Cross-framework verified

1

Measured runs

1

Independent sources

Other

Source platforms

17 days ago

Last verified

Performance

  1. llama.cpp · IQ4_XS Decode 90.66 · Prefill 438.77 ·
  2. llama.cpp · Q5_K_S (current)Decode 66.9 · Prefill 326.75 ·
  3. Ollama · Q2_K Decode 96.52 · Prefill 478.87 ·

Core figures

Decode (avg)
66.9 tok/s
Prefill (avg)
326.75 tok/s
TTFT (avg)
1.94 s
VRAM (avg)
24.78 GB
MTP acceptance rate
—
TTFB
0.54 GB
Power draw
116.8 W

Configuration

Member-level fields are taken from the most recent run

Model
gemma-4-12B-it
Quantization
Q5_K_S
Framework
llama.cpp
Version
v0.42
Context length
2048 tokens
Batch size
256
GPU layers
20
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)
24.78 GB
OS
macOS 26.1
Power draw
116.8 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

    v0.42 · macOS 26.1 · 2048 ctx

    66.9 tok/s

    Decode

    326.75 tok/s

    Prefill

    1.94 s

    TTFT

    24.78 GB

    VRAM

    —

    MTP

    116.8 W

    Power draw

    L2 batch seed 2026-10-19: llama.cpp on hw_id=103 running model_id=65 at Q5_K_S