Ceva announced on November 2 that Mobilint’s REGULUS edge processor is the first production deployment of its NeuPro NPU family. The chip combines NeuPro-Nano and SensPro processing with Mobilint’s architecture. The companies describe 10 TOPS at under 3 W and support for up to 8 GB of memory.
Production silicon is the milestone
Those are supplier specifications. End-to-end power and useful throughput depend on model precision, memory traffic and how work is divided among processing engines. A peak operations figure offers limited information about a particular application’s latency.
Workload placement matters
The immediate commercial route includes Mobilint’s MLD-R1 USB accelerator, which Ceva says is shipping. It offers a way to add local inference to existing systems without replacing the host computer. Integration still requires supported models, interfaces and software.
System measurements establish the benefit
The development could broaden access to compact edge AI. Representative workload tests should connect accuracy, latency and total system power, including data movement. Those measurements would show where the combined architecture delivers useful performance.
Image: Ceva / source announcement. Used for source-attributed editorial reporting; rights remain with the owner.
Sources & evidence
Sources describe the announcements and evidence available when this analysis was published. Company claims are attributed; plans and trials are not treated as sustained operation.
