NVIDIA announced on April 14 the release of Ising, the world’s first family of open-source AI models designed specifically to tackle the two engineering problems that have kept quantum computing from becoming actually useful: calibration and error correction. The full weights, training data, and tools are already live on GitHub, Hugging Face, and build.nvidia.com, available to researchers and developers worldwide at no cost.
The timing of the announcement was deliberate. NVIDIA chose World Quantum Day, an international initiative celebrated every April 14 by researchers across more than 65 countries, to make the release official through its own virtual NVIDIA Quantum Day event, a five-hour online program featuring sessions on the company’s quantum roadmap, AI-driven calibration, and real-time error correction.
The date itself carries meaning in the physics world: April 14, or 4.14, approximates the first digits of Planck’s constant, the fundamental value that governs quantum mechanics. NVIDIA used the moment to put a stake in the ground.
The name Ising also comes from physics history, a foundational mathematical model that dramatically simplified how scientists understood complex physical systems. The parallel is intentional. NVIDIA wants Ising to do the same thing for quantum hardware.
Why Quantum Computing is still stuck
Qubits, the basic unit of information in a quantum processor, are inherently unstable. They’re sensitive to temperature, electromagnetic interference, even vibration. The result is errors, a lot of them. Right now, the best quantum processors on the planet produce an error roughly once every thousand operations.
That might sound manageable, but for quantum computing to run real-world scientific and enterprise applications, that number needs to drop to one error in a trillion operations or less. The gap between where the technology is today and where it needs to be is enormous, and two specific problems are at the root of it.
The first is calibration. Keeping a quantum processor tuned and running correctly requires continuous adjustment to account for hardware imperfections and environmental noise. Currently, that process is slow, doesn’t scale, and relies heavily on human intervention.
The second is error correction decoding, the real-time process of catching and fixing qubit errors faster than they accumulate, which requires processing terabytes of measurement data thousands of times per second.
These two bottlenecks are exactly what Ising was built to attack.
What Ising actually does
The Ising family launches with two distinct model types. The first is Ising Calibration, a 35-billion-parameter vision-language model fine-tuned to read experimental measurements directly from a quantum processing unit and determine what adjustments need to be made to keep it running within spec.
When paired with an AI agent, it automates the entire calibration workflow, cutting a process that previously took days down to hours. The model was trained on data spanning multiple qubit modalities, including superconducting qubits, quantum dots, ions, neutral atoms, and electrons on helium.
The second is Ising Decoding, which comes in two variants of a 3D convolutional neural network, one with 0.9 million parameters optimized for speed, and one with 1.8 million parameters optimized for accuracy, both designed to perform real-time pre-decoding for quantum error correction.
NVIDIA benchmarked Ising Decoding against pyMatching, the open-source decoder most quantum research groups currently rely on, and the results are significant: 2.5 times faster and 3 times more accurate, while requiring ten times less training data to get there.
Sam Stanwyck, NVIDIA’s director of quantum product, explained the stakes clearly: a 2.5x speedup in decoding directly raises the ceiling on how many gate operations a quantum processor can sustain before its logical qubits collapse. In practical terms, the quantum computer can do more meaningful work before errors render the calculation useless.
Jensen Huang, NVIDIA’s founder and CEO, framed the release this way: “AI is essential to making quantum computing practical. With Ising, AI becomes the control plane, the operating system of quantum machines, transforming fragile qubits to scalable and reliable quantum-GPU systems.”
Open, integrated, and already moving markets
Ising integrates directly with NVIDIA’s CUDA-Q software platform for hybrid quantum-classical computing and with the NVQLink QPU-GPU hardware interconnect for real-time control and error correction. The models can also run locally on a researcher’s own systems, which matters for institutions that can’t afford to have proprietary quantum hardware data sitting on external infrastructure.
NVIDIA is also shipping a cookbook of quantum workflows and training data alongside NIM microservices, so developers can fine-tune Ising for their specific hardware configurations without needing deep machine learning expertise to get started.
Adoption is already wide. Among the institutions using Ising Calibration: Atom Computing, IonQ, Infleqtion, and Lawrence Berkeley National Laboratory’s Advanced Quantum Testbed. On the decoding side: the University of Chicago, Sandia National Laboratories, SEEQC, and IQM Quantum Computers.
Academic heavyweights including Harvard’s John A. Paulson School of Engineering and Applied Sciences, Cornell University, Fermi National Accelerator Laboratory, Academia Sinica, and the U.K. National Physical Laboratory are also on board.
The market took notice immediately. On the day of the announcement, quantum computing stocks surged: D-Wave jumped 10.3%, IonQ gained 13.3%, and Rigetti climbed 8.9%. That kind of reaction reflects how significant the industry considers NVIDIA’s entry into this space.
The Quantum Economic Development Consortium also released its State of the Global Quantum Industry 2026 report the same day, confirming the global quantum market reached $1.9 billion in 2025 and is forecast to grow at 30% annually.
The quantum computing market overall is projected to surpass $11 billion by 2030, according to analyst firm Resonance, but that growth depends entirely on solving exactly the problems Ising is targeting.
By open-sourcing these tools, NVIDIA is betting that putting the best calibration and decoding AI in as many hands as possible is the fastest path to making quantum hardware genuinely useful, and keeping itself at the center of the ecosystem in the process.
Do you think AI is the missing piece that finally makes quantum computing a reality? Tell us what you think in the comments, we’d love to hear your take!