descriptionMACOS ENVIRONMENT

macOS Platform Setup

Optimize your inference testing environment for Apple hardware. AIBench v2.0 provides native acceleration for Apple Silicon (M1/M2/M3/M4) and comprehensive support for legacy Intel architectures.

Global Installation

$pip install aibench-cli

Architecture Support

memory

Apple Silicon

Native acceleration utilizing unified memory architecture for M1, M2, M3, and M4 families. Yields maximum throughput with minimal thermal throttling.

check_circleRECOMMENDED

developer_board

Intel Architecture

Full compatibility for older Mac models via ONNX CPU execution providers. Performance scales with available cores and memory bandwidth.

Requires macOS 11.0+ (Big Sur)

Compute Backends

BACKEND ENGINE	APPLE SILICON	INTEL MAC
CoreML	check	remove
Metal Performance Shaders (MPS)	check	remove
ONNX CPU	check	check
PyTorch (CPU)	check	check
TFLite (CPU)	check	check

warningNote on Rosetta 2

If you are running a Python environment compiled for x86_64 on an Apple Silicon Mac, AIBench will execute via Rosetta 2. This operates exclusively on the CPU and will severely impact inference performance metrics. Always ensure you are using an arm64 compiled Python distribution (e.g., via Miniforge or native Homebrew) to unlock native hardware acceleration.