Top AI Product

Cognex shipped the In-Sight 6900 on April 28 — an industrial edge AI vision controller that sits next to a production line, runs AI models on incoming camera frames, and skips the PC entirely.

157 TOPS at the edge

Inside is an NVIDIA Jetson Thor-class module hitting 157 TOPS. Enough to run several high-resolution models in parallel — defect detection, OCR, semantic segmentation — without bouncing data back to a server. It supports transformer-based few-sample classification (10–20 training images), pixel-level semantic segmentation, and NVIDIA TensorRT for microsecond-scale timing sync with PLCs. Hardware is modular: cameras, lenses, and lighting all swap independently. CGNX moved on the news the same day.

API, PLC, and OneVision integration

This box is fully addressable from code. The In-Sight Vision Suite (ISVS) SDK plus Cognex’s OneVision platform expose REST APIs for training, deployment, and pulling inspection results. On the floor, EtherNet/IP, PROFINET, and TCP/IP socket commands handle remote triggers, parameter pushes, and verdicts. An AI agent can fire an inspection, read the verdict, and route it into a PLC workflow with no human in the loop. OneVision is the collaborative layer engineering teams use to train and ship models across plants.

You Might Also Like

• Z ais glm 5 1 Tops swe Bench pro at 58 4 Trained on Zero Nvidia Hardware
• Arduino Ventuno q Takes aim at Jetson Orin Nano With 40 Tops and an Stm32 Onboard
• Advantech mic 735 ai Inference System Puts Nvidia igx Thor Inside a Ruggedized amr Brain
• Sixfab Alpon x5 ai Runs 25 Tops on a Raspberry pi cm5 Without Nvidia Inside
• Ricursive Intelligence Just Raised 335m in 4 Months and i Think Nvidia Should be Paying Attention