Arao AI
Back to search results

gemma-4-E2B-it · Q4_K_M

Start a discussion
Apple M2 ProvLLM v0.49L2 Cross-framework verified

This page aggregates 1 real-world runs of gemma-4-E2B-it (Q4_K_M) on Apple M2 Pro with vLLM, contributed by 1 independent source platforms; metrics are averages of published measurements.

128.43 tok/s

Decode

Decode speed

487.16 tok/s

Prefill

Prefill speed

0.75 s

TTFT

Time to first token

5.54 GB

VRAM

VRAM usage

L2 Cross-framework verified

1

Measured runs

1

Independent sources

Other

Source platforms

2 days ago

Last verified

Performance

  1. vLLM · Q4_K_M (current)Decode 128.43 · Prefill 487.16 ·
  2. Ollama · IQ2_XS Decode 158.03 · Prefill 828.05 ·
  3. Ollama · Q3_K_S Decode 140.03 · Prefill 568.6 ·

Core figures

Decode (avg)
128.43 tok/s
Prefill (avg)
487.16 tok/s
TTFT (avg)
0.75 s
VRAM (avg)
5.54 GB
MTP acceptance rate
—
TTFB
1.03 GB
Power draw
39.1 W

Configuration

Member-level fields are taken from the most recent run

Model
gemma-4-E2B-it
Quantization
Q4_K_M
Framework
vLLM
Version
v0.49
Context length
32768 tokens
Batch size
32
GPU layers
999
Flash Attention
Off

Hardware

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

GPU
Apple M2 Pro
Measured VRAM (avg)
5.54 GB
OS
macOS 15.0
Driver
Apple M2 Pro
CUDA
N/A
Power draw
39.1 W

Sources & evidence

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

  1. L2 Cross-framework verifiedOtherOriginal link Verified on 2026-10-16

    v0.49 · macOS 15.0 · CUDA N/A · 32768 ctx

    128.43 tok/s

    Decode

    487.16 tok/s

    Prefill

    0.75 s

    TTFT

    5.54 GB

    VRAM

    —

    MTP

    39.1 W

    Power draw

    L2 batch seed 2026-10-16: vLLM on hw_id=102 running model_id=66 at Q4_K_M