Ambiq has launched two new edge AI runtime solutions, HeliosRT and HeliosAOT, on July 22, 2025, to improve performance and energy efficiency for AI models on Apollo SoCs.

Both solutions are powered by Ambiq’s patented Sub-threshold Power Optimized Technology (SPOT), which has been implemented in over 270 million devices globally, ensuring significant performance benefits.

These solutions are designed to help deploy sophisticated AI models in battery-powered edge devices that require high performance within strict power budgets.

HeliosRT is available in beta via the neuralSPOT SDK with a planned general release in Q3 2025, while HeliosAOT is in technical preview with broader availability expected in Q4 2025.

HeliosAOT, an ahead-of-time compiler, converts TensorFlow Lite models into embedded C code, reducing memory footprint by 15-50% and offering greater deployment flexibility.

HeliosRT, an energy-optimized version of LiteRT, features custom AI kernels that can improve inference speed and power efficiency by up to three times, especially on the Apollo510 chip.

These tools come with comprehensive documentation, examples, and engineering support to assist developers in leveraging their capabilities.

HeliosAOT provides a 15-50% reduction in memory footprint compared to traditional deployment methods, making it ideal for resource-constrained systems.

Ambiq aims to empower developers to unlock the full potential of Apollo SoCs, enabling AI deployment across various applications like digital health, smart homes, and industrial automation.

The solutions are specifically optimized for Ambiq’s Apollo family of SoCs, addressing the demands of edge computing environments where power efficiency is critical.