NVIDIA, on World Quantum Day, has announced Ising, a family of open-source AI models designed to improve quantum computing systems. The company said the models will help researchers and enterprises build quantum processors capable of running useful applications.
The AI giant released two Ising models: Calibration and Decoding.
Ising Calibration uses a vision language model to interpret quantum processor data and automate tuning. This reduces calibration time from days to hours, according to the company.
Ising Decoding includes two neural network models designed for speed and accuracy in quantum error correction. These models process large volumes of data in real-time and improve system performance when combined with existing decoders.
During a briefing, Sam Stanwyck, Director for Quantum Product at NVIDIA, outlined how AI can address key limitations in current quantum hardware.
The Ising models focus on quantum processor calibration and error correction—two problems caused by noisy qubits. This noise limits reliability and scale in current systems.
“Qubits are noisy, and the way to manage that noise at the scale we need is with AI models,” he said, adding that AI can act as the “control plane for quantum hardware.”
NVIDIA said its models can deliver up to 2.5x faster performance and 3x higher accuracy in decoding compared to existing methods.
The company also positioned its GPUs as central to solving quantum computing challenges, leveraging the AI models to process large volumes of quantum data in real-time and manage system noise.
“AI is essential to making quantum computing practical,” said NVIDIA chief Jensen Huang in a statement. “With Ising, AI becomes the control plane, the operating system of quantum machines, transforming fragile qubits to scalable and reliable quantum-GPU systems.”
Stanwyck also responded to AIM’s query on how Ising integrates itself into the Indian ecosystem. He said developers can still work on quantum problems using simulation tools.
He noted that NVIDIA’s libraries can simulate both noisy and noiseless quantum systems, enabling research without physical access to hardware. Through this, he addressed that even if an ecosystem doesn’t have mature hardware, the companies can still use Ising models.
Several research institutions and enterprises have already started using the models. These include academic labs, national laboratories, and quantum hardware companies, with some users including Atom Computing, IonQ, IQM Quantum Computers, Fermi National Accelerator Laboratory and Cornell University, among others.
The models integrate with NVIDIA’s existing quantum platform, including CUDA-Q and NVQLink, where Ising adds an AI layer to this stack for system control and integration.
On timelines for commercial use, Stanwyck said the models have already begun improving quantum systems. However, he did not provide a clear timeline for real-world applications.
“Our goal is to try to move those [roadmaps] in as much as possible,” he said, referring to the roadmaps publicly announced by tech companies such as Google, IBM, and Microsoft.
The company noted that the quantum computing market could exceed $11 billion by 2030 and that progress depends on solving challenges such as error correction and scalability.
NVIDIA is releasing the models as open-source to allow developers to customise them for specific hardware and use cases. It also offers workflow cookbooks, training data, and NIM microservices to help developers fine-tune models for specific hardware. The models can run locally, allowing teams to protect proprietary data.
Ising now joins NVIDIA’s wider portfolio of open models, including NVIDIA Nemotron, Cosmos, Alpamayo, Isaac GR00T, and BioNeMo. These are available on platforms such as GitHub and Hugging Face.
Your reaction
Discussion
What's your take on this story?
No comments yet — be the first to share your view.