MediaTek’s Genio 720 & 520: Smarter, faster AI for the next-gen IoT

MediaTek has raised the bar for edge AI computing with the launch of Genio 720 and Genio 520, two cutting-edge IoT platforms designed to power the next generation of smart devices. Announced at Embedded World, these AI-driven platforms bring robust generative AI capabilities, high-speed connectivity, and enhanced performance to smart home, retail, industrial, and commercial IoT applications.

Built on a 6nm process node, both platforms feature an octa-core CPU (2x Arm Cortex-A78 + 6x Cortex-A55). It will boost performance without draining power, perfect for battery-operated gadgets, ensuring an optimal balance of power efficiency and performance.

The 8th-gen Neural Processing Unit (NPU) delivers up to 10 TOPS of computing power, providing full hardware acceleration for AI workloads, including transformer and convolutional neural network (CNN) models.

One of the key highlights is support for large language models (LLMs) such as Llama, Gemini, Phi, and DeepSeek, enabling faster, on-device generative AI tasks without relying on cloud processing.

With up to 16GB LPDDR5 memory, these platforms efficiently handle multi-modal AI applications, ensuring seamless HMI interactions and advanced multimedia processing. This means that these platforms can run large AI models (like ChatGPT-style assistants) directly on devices, ensuring user privacy.

Developers will benefit from OSM (Open Standard Module) reference designs, which streamline product development and accelerate time-to-market. MediaTek’s partnerships will bring OSM-based Genio 720/520 solutions to market in H2 2025, making it easier for companies to integrate AI-powered IoT solutions.

MediaTek’s plug-and-play design tools mean companies can launch AI-powered devices faster than ever.

With 4K/5K ultrawide display support, dual ISP for high-resolution cameras, Wi-Fi 6/6E, and expandable Wi-Fi 7/5G RedCap compatibility, the Genio 720 and 520 set a new standard for AI-enhanced, power-efficient IoT innovation. Sampling for customers begins in Q2 2025.