Nvidia has launched a family of open source AI models dubbed Ising to support the development of quantum chips.

Named after a mathematical model that helped to simplify the understanding of complex physical systems, Ising comprises two model domains, calibration and decoding.

Quantum calibration is the continuous tuning of control parameters in order to keep quantum processors operational, a task that is often carried out by quantum physicists or simple automated algorithms. Decoding algorithms, by comparison, are used to tackle quantum error correction – techniques that protect quantum information from errors that arise from noisy qubits.

“That noise is the fundamental bottleneck standing between today's quantum hardware and useful applications,” said Sam Stanwyck, director of quantum product at Nvidia, ahead of the launch.

Nvidia’s Ising Calibration tool is a vision language model that, according to the chip giant, can interpret and react to measurements from quantum processors, allowing AI agents to “automate continuous calibration,” thus reducing the time needed from days to hours.

Meanwhile, Ising Decoding provides two variants of a 3D convolutional neural network model that can be optimized for either speed or accuracy. This allows the model to perform real-time decoding for quantum error correction, with Nvidia claiming that these models are up to 2.5x faster and 3x more accurate than pyMatching, the current open source industry standard.

In addition to the “state-of-the-art customizable” models, Stanwyck said Nvidia is also shipping a cookbook with fine-tuning, quantization, and inference workflows, alongside “recipes” for how to integrate the tools into agentic workflows, open research papers, and benchmarking data. This will allow developers to optimize the models for their specific hardware and noise characteristics, he added.

Ising is compatible with Nvidia’s CUDA-Q software platform for hybrid quantum-classical computing and also integrates with the company’s NVQLink QPU-GPU hardware interconnect.

Nvidia said Ising is already being used by a number of quantum computing companies and research labs across the globe. This includes Atom Computing, Academia Sinica, Fermi National Accelerator Laboratory, Infleqtion, IonQ, IQM Quantum Computers, Lawrence Berkeley National Laboratory’s Advanced Quantum Testbed, Q-CTRL, and the UK National Physical Laboratory, all of which have deployed Ising Calibration.

Furthermore, Ising Decoding has been deployed by Cornell University, Infleqtion, IQM Quantum Computers, Quantum Elements, Sandia National Laboratories, University of Chicago, University of Southern California, and Yonsei University.

“The very best quantum processors make an error about once in every 1000 operations, which is amazing, but to become useful accelerators for scientific and enterprise-valuable problems, that number needs to become one in a trillion or even less,” Stanwyck said. “The good news is that AI can be the answer for how you manage that noise at scale, and it has the potential to enable very rapid progress in closing that gap.”

He added: “Ising delivers the world's best performance on both calibration and error correction decoding, and this is the path to quantum GPU supercomputing, which is a quantum accelerator, integrating with the GPU supercomputer to solve valuable problems. So, the way to think about this is, AI is becoming the control plane for quantum hardware.”

The Nvidia Ising models, data, and frameworks are available on GitHub, Hugging Face, and Nvidia’s Build developer hub.