Arao AI
Back to search results

Qwen3-4B · FP16

Start a discussion
Apple M1 ProStrata v0.51L2 Cross-framework verified

This page aggregates 1 real-world runs of Qwen3-4B (FP16) on Apple M1 Pro with Strata, contributed by 1 independent source platforms; metrics are averages of published measurements.

58.73 tok/s

Decode

Decode speed

267.72 tok/s

Prefill

Prefill speed

2.27 s

TTFT

Time to first token

17.64 GB

VRAM

VRAM usage

L2 Cross-framework verified

1

Measured runs

1

Independent sources

Other

Source platforms

2 days ago

Last verified

Performance

  1. Ollama · IQ3_XS Decode 98.64 · Prefill 472.33 ·
  2. Strata · FP16 (current)Decode 58.73 · Prefill 267.72 ·

Core figures

Decode (avg)
58.73 tok/s
Prefill (avg)
267.72 tok/s
TTFT (avg)
2.27 s
VRAM (avg)
17.64 GB
MTP acceptance rate
0.63
TTFB
0.47 GB
Power draw
47.6 W

Configuration

Member-level fields are taken from the most recent run

Model
Qwen3-4B
Quantization
FP16
Framework
Strata
Version
v0.51
Context length
16384 tokens
Batch size
512
GPU layers
12
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)
17.64 GB
OS
macOS 15.0
Driver
Apple M1 Pro
CUDA
N/A
Power draw
47.6 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

    v0.51 · macOS 15.0 · CUDA N/A · 16384 ctx

    58.73 tok/s

    Decode

    267.72 tok/s

    Prefill

    2.27 s

    TTFT

    17.64 GB

    VRAM

    0.63

    MTP

    47.6 W

    Power draw

    L2 batch seed 2026-10-16: Strata on hw_id=100 running model_id=54 at FP16