Nvidia on the 16th unveiled Nvidia Ising, the world's first open-source quantum AI model family that it said will speed the practical commercialization of Quantum Computing. With the release, research institutions and corporations are expected to dramatically accelerate development of quantum processors capable of running practical applications.
To implement large-scale quantum applications, technical leaps in quantum processor calibration and quantum error correction are essential. Nvidia defined AI as the key technology to turn today's quantum processors into reliable, large-scale computers. The newly released open-source models help developers build high-performance AI while maintaining control over their data and infrastructure.
The Ising family takes its name from a mathematical model that simplifies complex physical systems and provides high-performance AI tools for error correction and calibration, the core challenges in building hybrid quantum-classical systems. In particular, in quantum error correction decoding, it delivers up to 2.5 times faster performance and three times higher accuracy than pymatching, the existing industry-standard tool, enabling researchers to solve more complex problems.
Jensen Huang, Nvidia CEO, said, "AI is essential to make Quantum Computing practical," adding, "With Ising, AI serves as the operating system of quantum machines as the control layer, transforming fragile Qubit into scalable and reliable quantum-GPU systems."
The family includes two primary models. Ising Calibration uses a vision-language model (VLM) to quickly interpret quantum processor measurement results and automate calibration tasks. This can cut jobs that used to take days down to just a few hours. Ising Decoding, based on a 3D convolutional neural network (CNN), performs real-time error correction and comes in two versions optimized separately for speed and accuracy.
Leading corporations and institutions worldwide are adopting Ising to advance Quantum Computing development. Ising Calibration is in use at Atom Computing, Academia Sinica in Taiwan, Iroqu, Conductor Quantum, Fermi National Accelerator Laboratory, Harvard School of Engineering, Infleqtion, IonQ, IQM Quantum Computers, the Advanced Quantum Testbed at Lawrence Berkeley National Laboratory, Q-CTRL, and the United Kingdom's National Physical Laboratory, among others.
Ising Decoding has been adopted by Cornell University, Eddencode, Infleqtion, IQM Quantum Computers, Quantum Elements, Sandia National Laboratories, SecuC, the University of California San Diego and Santa Barbara, the University of Chicago, the University of Southern California, and, in Korea, Yonsei University.
Nvidia also provides a cookbook containing training data and workflows so developers can fine-tune the models for their hardware architectures. Ising is integrated with the Nvidia CUDA-Q software platform and NVQ Link hardware, completing a comprehensive toolset for building next-generation quantum supercomputers. The Quantum Computing market is forecast to surpass $11 billion in 2030, and the Ising models are available via GitHub and Hugging Face.