Ceva and embedUR have launched a dedicated instance of the ModelNova platform for the NeuPro NPU family, providing developer‑ready AI models for both the ultra‑low‑power NeuPro‑Nano and the high‑performance NeuPro‑M processors.
The ModelNova platform, built by embedUR, offers a curated library of models optimized with TensorFlow Lite Micro and integrated with Ceva’s NeuPro Studio SDK. Models include object detection, face recognition, pose estimation, audio denoising, keyword spotting, and anomaly detection, each accompanied by performance metrics, execution scripts, and dataset descriptions.
The dedicated instance hosts a selection of models from both Ceva and embedUR, enabling rapid prototyping and deployment across wearables, hearables, smart‑home devices, industrial automation, robotics, smart cameras, AR/VR, and future transformer‑based architectures such as large language models and generative AI.
Ceva’s NeuPro‑Nano delivers 10 GOPS to 200 GOPS per core, while the NeuPro‑M supports 4 TOPS to 256 TOPS per core, scaling to over 1,200 TOPS in multi‑core configurations. The platform’s immediate availability is expected to reduce time‑to‑market for Ceva customers and expand adoption of its NPUs across a broader range of devices.
embedUR’s Eric Smiley said the platform will empower developers to bring AI to devices faster. Ceva’s Elia Shenberger added that the partnership will accelerate edge AI innovation across a wide range of applications.
The collaboration aligns with the growing “Physical AI” trend, embedding intelligence directly into devices for autonomous operation without cloud connectivity, addressing critical needs for low latency, privacy, and reliability.
The content on BeyondSPX is for informational purposes only and should not be construed as financial or investment advice. We are not financial advisors. Consult with a qualified professional before making any investment decisions. Any actions you take based on information from this site are solely at your own risk.