Nvidia Open-Sources ISING to Speed Quantum Chip Calibration and Error Correction

Overview

• Nvidia released ISING as an open-source model suite with code, training data, and workflows on GitHub, Hugging Face, and its website.
• ISING plugs into the company’s CUDA-Q software, NVQLink links between quantum and graphics chips, and NIM microservices for easy tuning and local deployment that keeps proprietary data on site.
• The Calibration model uses a vision–language system to read hardware measurements and automate setup, which Nvidia says can cut tuning time from days to hours.
• The Decoding model uses 3D neural networks for real-time error correction and offers speed or accuracy variants, with Nvidia reporting up to 2.5× faster and up to 3× more accurate than the common pyMatching baseline.
• Early users span universities, labs, and firms, including Cornell, UC San Diego, the University of Chicago, Yonsei, IonQ, IQM, and Sandia National Laboratories.