Arao AI
Back to search results

Qwen3-4B-Thinking-2507 · Q2_K

Start a discussion
Apple M1 ProStrata v0.30L2 Cross-framework verified

This page aggregates 1 real-world runs of Qwen3-4B-Thinking-2507 (Q2_K) on Apple M1 Pro with Strata, contributed by 1 independent source platforms; metrics are averages of published measurements.

151.41 tok/s

Decode

Decode speed

638.55 tok/s

Prefill

Prefill speed

0.71 s

TTFT

Time to first token

6.72 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 · IQ3_S Decode 94.49 · Prefill 403.53 ·
  2. Strata · IQ1_S Decode 163.46 · Prefill 649.89 ·
  3. Strata · Q2_K (current)Decode 151.41 · Prefill 638.55 ·

Core figures

Decode (avg)
151.41 tok/s
Prefill (avg)
638.55 tok/s
TTFT (avg)
0.71 s
VRAM (avg)
6.72 GB
MTP acceptance rate
0.58
TTFB
1.21 GB
Power draw
24 W

Configuration

Member-level fields are taken from the most recent run

Model
Qwen3-4B-Thinking-2507
Quantization
Q2_K
Framework
Strata
Version
v0.30
Context length
16384 tokens
Batch size
1024
GPU layers
5
Flash Attention
On

Hardware

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

GPU
Apple M1 Pro
Measured VRAM (avg)
6.72 GB
OS
macOS 15.0
Driver
Apple M1 Pro
CUDA
N/A
Power draw
24 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

    v0.30 · macOS 15.0 · CUDA N/A · 16384 ctx

    151.41 tok/s

    Decode

    638.55 tok/s

    Prefill

    0.71 s

    TTFT

    6.72 GB

    VRAM

    0.58

    MTP

    24 W

    Power draw

    L2 batch seed 2026-10-16: Strata on hw_id=100 running model_id=57 at Q2_K