Arao AI
Back to search results

gemma-4-26B-A4B-it · Q3_K_S

Start a discussion
Apple M5 Max128GBvLLM v0.25L2 Cross-framework verified

This page aggregates 1 real-world runs of gemma-4-26B-A4B-it (Q3_K_S) on Apple M5 Max with vLLM, contributed by 1 independent source platforms; metrics are averages of published measurements.

174.38 tok/s

Decode

Decode speed

573.36 tok/s

Prefill

Prefill speed

1.98 s

TTFT

Time to first token

37.59 GB

VRAM

VRAM usage

L2 Cross-framework verified

1

Measured runs

1

Independent sources

Other

Source platforms

5 days ago

Last verified

Performance

  1. vLLM · Q3_K_S (current)Decode 174.38 · Prefill 573.36 ·
  2. vLLM · Q4_K_S Decode 136.41 · Prefill 547.6 ·
  3. Strata · Q2_K Decode 228.98 · Prefill 895.33 ·
  4. Strata · Q8_0 Decode 122.58 · Prefill 429.76 ·

Core figures

Decode (avg)
174.38 tok/s
Prefill (avg)
573.36 tok/s
TTFT (avg)
1.98 s
VRAM (avg)
37.59 GB
MTP acceptance rate
—
TTFB
1.4 GB
Power draw
42.3 W

Configuration

Member-level fields are taken from the most recent run

Model
gemma-4-26B-A4B-it
Quantization
Q3_K_S
Framework
vLLM
Version
v0.25
Context length
16384 tokens
Batch size
8
GPU layers
999
Flash Attention
On

Hardware

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

GPU
Apple M5 Max
Nominal VRAM
128 GB
Measured VRAM (avg)
37.59 GB
OS
macOS 15.0
Driver
Apple M5 Max
CUDA
N/A
Power draw
42.3 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

    v0.25 · macOS 15.0 · CUDA N/A · 16384 ctx

    174.38 tok/s

    Decode

    573.36 tok/s

    Prefill

    1.98 s

    TTFT

    37.59 GB

    VRAM

    —

    MTP

    42.3 W

    Power draw

    L2 batch seed 2026-10-13: vLLM on hw_id=28 running model_id=64 at Q3_K_S