Ceva and Mobilint unveil REGULUS, an on-device AI processor built on Mobilint’s REGULUS and SensPro/NPU mix, delivering real-time sensing, interpretation, and response for power-constrained edge devices.

The MLD-R1 platform supports major AI ecosystems, including PyTorch, ONNX, Hugging Face, TensorFlow, TFLite, and Keras, enabling broad model compatibility.

REGULUS achieves up to 10 TOPS of AI performance at under 3W, with support for up to 8GB RAM, targeting robotics, drones, machine vision, smart cameras, and IoT applications.

Market context from Pluang as of 2026-11-02, noting movements in US tech stocks, though these fluctuations are peripheral to the Ceva-Mobilint announcement.

Ceva has shipped over 22 billion devices and serves more than 400 customers, highlighting its role in accelerating development and reducing risk for intelligent edge projects.

Ceva positions its AI fabric portfolio to connect, sense, and infer across 5G, Bluetooth, Wi-Fi, and UWB with low-power, scalable edge AI IP.

MLD-R1 enables real-time pedestrian detection, collision-prevention alerts, onboard vision for AGVs/AMRs, restricted-zone intrusion detection, and worker safety monitoring at sub-10W power draw.

The collaboration enhances Physical AI by enabling efficient sensing and locally running AI workloads on edge devices.

Mobilint’s SDK enables compilation of PyTorch, TensorFlow, and ONNX models, with over 490 validated architectures across CNN, RNN, and transformer types on the hardware.

Mobilint’s Yocto Linux SDK includes GStreamer and OpenCV with camera/video/display drivers, enabling streamlined development on the REGULUS platform.

Mobilint is a Seoul-based fabless AI semiconductor company focused on high-performance, energy-efficient NPUs and AI SoCs for edge and on-device AI across multiple industries.

Mobilint’s REGULUS is part of a broader partnership with Ceva, leveraging NeuPro NPU and SensPro DSP as part of an AI fabric for smart edge deployments.