In a world where the race towards quantum computing is accelerating, the American company NVIDIA announced the launch of the Ising Models family, which is the first family of open source artificial intelligence models specifically designed to accelerate the construction of quantum processors capable of running real applications.
The models are named after a famous mathematical model used in statistical physics to describe interactions between atoms, but Nvidia – whose main activity is focused on graphics processing units that support artificial intelligence – has turned it into an artificial intelligence tool with modern digital vision.
The quantum computing market is expected to exceed US$11 billion in 2030, according to research firm Resonance, but this growth depends largely on continued progress in addressing fundamental engineering challenges, such as quantum error correction.
What are Nvidia’s iZing models?
In recent years, quantum computing has faced two major challenges to find useful applications: quantum error correction and real-time calibration, and Ising’s models directly target both, opening the door to hybrid systems that combine quantum and classical computers.
“AI is essential to making quantum computing practical,” said Nvidia CEO Jensen Huang. “With Izing models, AI becomes the operating system for quantum machines.”
The Izing family includes two fully open source models for real-time debugging and calibration, available via Hugging Face and GitHub, with training frameworks and complete datasets licensed by NVIDIA that allow for local modification and training without data sharing.
The Ising Calibration model interprets the results from quantum processors and proposes automatic calibration actions through AI agents to automate QPU calibration tasks, reducing the required runtime from days to hours.
The qubits in quantum computers need precise and continuous calibration to ensure that they work in the desired harmony, and Ising Calibration predicts deviations in the performance of the qubits and corrects them in real time before they lead to the failure of the calculations.
This language vision model contains 35 billion parameters, is trained on multimodal data from different types of quantum processing units, and outperforms other large models on a new QCalEval test that measures quantum calibration performance.
Traditional calibration methods rely on physicists or pre-defined calibration processes to fine-tune systems before each boot, while NVIDIA designed the Ising Calibration model to continuously recalibrate as hardware changes over time, allowing systems to scale much further.
While the Ising Decoding model comes as a complementary layer and is not a replacement for existing methods, it is available in two 3D neural network models, one optimized for speed and the other for accuracy.
This model acts as a primary error correction unit that uses neural networks to process data or error signals derived from qubit measurements and correct a large portion of errors before passing the processed data to traditional algorithms.
While calibration is concerned with prevention, the error correction model is concerned with treatment. This model monitors the output of a quantum computer, accurately identifies noise and separates it from real data, thanks to its training on billions of scenarios of error patterns.
This hybrid approach aims to improve speed and accuracy while maintaining compatibility with existing debugging pipelines, and integrate with the traditional quantum hybrid computing platform CUDA-Q and NVQLink technology.
The role of models in handling errors in quantum computers
The best quantum processors make about one error in every thousand operations, and this number would need to drop to one in a trillion or even less for these processors to become useful in solving important scientific and commercial problems.
Here comes the role of quantum calibration, which is defined as the process of constantly adjusting control parameters to ensure the continued operation of quantum processors, a task often carried out by quantum physicists or simple automated algorithms.
As for error correction algorithms, they are used to address quantum errors through advanced techniques that protect quantum information from errors caused by qubit noise by arranging the physical qubit in a two-dimensional network and repeating the information across it, which forms the protected logical qubit.
This process requires monitoring millions of measurements in real time and correcting the error faster than the error rate, i.e. within millionths of a second, while the precision of monitoring and error correction reduces the number of physical qubits needed to create a logical qubit.
Here the strength of the Ising De Coding model emerges, which, according to official NVIDIA tests, provides 2.5 times faster performance and 3 times higher accuracy compared to the current open source standard, with 10 times less need for training data.
Competition drives innovation, with Nvidia accelerating the path to adoption via AI, while IBM builds the physical foundation one step at a time, with collaboration possible, as Nvidia supports IBM’s quantum processing units.
Geopolitical competition in the quantum race
Once just a matter of science, quantum computing has become a pillar of national security and a tool for enforcing national sovereignty, as countries with the ability to break traditional encryption or discover new chemicals through quantum simulation will lead the twenty-first century.
Through IBM and Nvidia, the United States seeks to maintain its leadership, create an open system that attracts global talent, and set industry standards.
While China is focusing intensely on quantum communications and physical devices, trying to build an impenetrable quantum internet, this puts Ising’s models in direct opposition to Chinese research into home-grown error correction based on artificial intelligence.
While European Union countries find NVIDIA’s open models an opportunity to reduce the gap without the need for huge investments in building giant quantum computers, which enhances the ability of European startups in the fields of biochemistry and finance.
In conclusion, the launch of Izing models represents an announcement that quantum computing is no longer confined to physics laboratories, in addition to heralding the beginning of a new era in which artificial intelligence becomes an essential partner in bringing about the quantum revolution, which is no longer a distant dream.
While IBM is racing against time to achieve quantum benefit, Nvidia is providing the tools that make this progress scalable globally, as the future is an open, intelligent hybrid system that solves humanity’s major problems, not a single quantum computer.
