Arm Edge AI Libraries, Tools & Frameworks

Explore Edge AI Libraries, Tools, and Frameworks on Arm

Find the libraries, compilers, evaluation kits, and reference platforms you need to develop, optimize, and deploy edge AI on Arm devices.

Assess
Build
Optimize

ML embedded evaluation kit

A turnkey evaluation kit to run and benchmark ML workloads on Arm Cortex-M microcontrollers with Ethos-U NPUs.

FVP documentation

Guide to simulating Arm-based systems using Fixed Virtual Platforms, enabling testing and validation without physical hardware.

Hugging Face hub

Curated set of ML models optimized for Arm Cortex-M and Cortex-A edge devices across NLP, vision, and audio domains.

FreeRTOS Corstone reference

Official FreeRTOS examples for Corstone-3xx platforms including Cortex-M85, Ethos-U85/U65, and Mali-C55 integrations.

Arm ML model zoo

Repository of machine learning models and examples optimized for Arm architectures across devices and workloads.

Keil MDK IDE

Comprehensive IDE for Arm Cortex-M development featuring code generation, debugging, and real-time analysis tools.

Compilers

Explore Arm compilers and libraries designed to help you get the best performance, efficiency, and portability from your code.

MLEK CMSIS examples

Example ML applications using CMSIS-NN and Ethos-U in CMSIS-Pack format to demonstrate end-to-end embedded AI workflows.
Documentation

Corstone FVPs

Prebuilt Fixed Virtual Platforms for Corstone-based designs, ideal for early software development and hardware validation.

FVPs on macOS

Community guide for running Arm Fixed Virtual Platforms natively on macOS and Apple Silicon systems.

Corstone-300

Platform integrating Cortex-M55 and Ethos-U55 for energy-efficient ML and signal processing at the edge.

Corstone-320

Reference platform combining Cortex-M85, Ethos-U85, and Mali-C55 ISP to accelerate complex edge and vision workloads.

Corstone-315

Scalable IoT and vision platform featuring Cortex-M85, Ethos-U65, and Mali-C55 ISP for efficient embedded ML deployments.

Corstone-310

Combines Cortex-M85 and Ethos-U55 to deliver high-efficiency ML compute for embedded and edge applications.

Corstone-300 (MPS3)

Zephyr BSP for Cortex-M55 and Ethos-U55 via MPS3, supporting real-time and embedded ML prototyping.

Corstone-320 / 315 (MPS4)

Zephyr support for Cortex-M85 and Ethos-U55/U65 via MPS4 boards, enabling flexible real-time development.

Corstone-1000

Secure SoC platform integrating cloud-native capabilities for running trusted firmware and IoT workloads on Arm-based systems.

Visual Studio Code

Lightweight, extensible code editor supporting Arm toolchains, extensions, and cross-platform embedded development.

Jupyter Notebook

Interactive coding environment ideal for developing, testing, and visualizing ML or data workflows on Arm.

CMSIS-NN

Optimized neural network kernels for Arm Cortex-M that improve performance and reduce memory footprint. Ideal for deploying efficient ML inference on microcontrollers.
Documentation

CMSIS-DSP

High-performance DSP library for Cortex-M processors, providing signal processing functions for audio, sensor, and ML applications optimized for embedded systems.

Ethos-U Vela compiler

Optimizes LiteRT models for Ethos-U NPUs (U55/U65/U85), converting them into efficient NPU-friendly formats for edge ML acceleration.
Documentation

TOSA (Tensor Operator Set Architecture)

An open intermediate representation for portable neural network graph optimization and consistent operator behavior across ML frameworks.