Arao AI
Back to search results

gemma-4-31B-it · Q8_0

Start a discussion
Apple M5 Ultra512GBllama.cpp v0.38L2 Cross-framework verified

This page aggregates 1 real-world runs of gemma-4-31B-it (Q8_0) on Apple M5 Ultra with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.

37.99 tok/s

Decode

Decode speed

165.65 tok/s

Prefill

Prefill speed

2.29 s

TTFT

Time to first token

76.97 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 · Q3_K_L Decode 65.89 · Prefill 258.39 ·
  2. llama.cpp · Q8_0 (current)Decode 37.99 · Prefill 165.65 ·
  3. Ollama · Q2_K Decode 75.92 · Prefill 280.65 ·
  4. Ollama · Q8_0 Decode 38.85 · Prefill 140.28 ·

Core figures

Decode (avg)
37.99 tok/s
Prefill (avg)
165.65 tok/s
TTFT (avg)
2.29 s
VRAM (avg)
76.97 GB
MTP acceptance rate
—
TTFB
0.3 GB
Power draw
112.4 W

Configuration

Member-level fields are taken from the most recent run

Model
gemma-4-31B-it
Quantization
Q8_0
Framework
llama.cpp
Version
v0.38
Context length
32768 tokens
Batch size
512
GPU layers
28
Flash Attention
Off

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)
76.97 GB
OS
macOS 26.1
Power draw
112.4 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

    v0.38 · macOS 26.1 · 32768 ctx

    37.99 tok/s

    Decode

    165.65 tok/s

    Prefill

    2.29 s

    TTFT

    76.97 GB

    VRAM

    —

    MTP

    112.4 W

    Power draw

    L2 batch seed 2026-10-19: llama.cpp on hw_id=29 running model_id=63 at Q8_0