NVIDIA announces the world’s first “Ising model” built for quantum AI

NVIDIA (EDGES) has released the world’s first open-source model designed specifically for quantum AI, “NVIDIA Ising.” The project aims to address calibration and error-correction challenges that quantum processors face during scaling. Through an open-source format, this series of models provides developers with high-performance tools, helping research institutions and enterprises optimize the operating efficiency of quantum bits.

NVIDIA Ising shortens the calibration timeline

The operational stability of a quantum processor depends heavily on precise calibration procedures. NVIDIA Ising (Ising Calibration) models use visual language model technology, enabling fast interpretation of measurement data from quantum processors and taking action. This technology allows AI agents to automatically carry out ongoing calibration tasks, shortening traditional procedures to just a few hours. Multiple research organizations—including Taiwan’s Academia Sinica, the Fermi National Accelerator Laboratory, and Harvard University—have begun adopting this technology to improve the efficiency of automated maintenance for quantum hardware and reduce labor costs.

Convolutional neural network acceleration for error correction decoding

NVIDIA founder and CEO Jensen Huang said that AI is crucial to making quantum computing practical. With Ising, AI becomes the operating system for controlling quantum machines, turning fragile quantum bits into a scalable and reliable quantum GPU system.

Quantum Error Correction is the technical threshold for enabling large-scale computing. NVIDIA Ising decoding (Ising Decoding) provides two variations of 3D convolutional neural network (3D CNN) models, optimized respectively for processing speed and computing accuracy. Data show that the Ising decoding models’ processing speed is 2.5 times higher than the current industry open-source standard pyMatching, and accuracy is up to 3 times better. By producing error signals through real-time decoding of quantum bits, researchers can maintain the logical correctness of computation. To date, institutions including Cornell University and IonQ have deployed this model, exploring more complex quantum algorithm applications.

According to forecasts by market research firm Resonance, the global quantum computing market is expected to reach $11 billion by 2030, with growth momentum mainly driven by breakthroughs in engineering challenges. The NVIDIA Ising series not only provides models and training data, but also integrates with the NVIDIA CUDA-Q software platform and NVIDIA NVQLink hardware interconnect technology. This hybrid architecture design enables researchers to run models on local systems, ensuring the privacy of experimental data. In addition, through NVIDIA NIM microservices, developers can fine-tune for specific hardware architectures, helping existing quantum bits evolve into scalable quantum computing systems.

Ising open-source model for open download

The Ising model series name is derived from a classic mathematical model that simplifies research into complex physical systems, reflecting the design philosophy of simplifying the quantum control plane. Currently, NVIDIA Ising models are available for download on GitHub, Hugging Face, and NVIDIA’s official platform, joining NVIDIA’s open model product lineup. This lineup includes NVIDIA Nemotron for agents, NVIDIA Cosmos for embodied AI, NVIDIA Alpamayo for autonomous vehicles, NVIDIA Isaac GR00T for robotics, and NVIDIA BioNeMo for biomedical research.

By positioning AI as the “control plane” or operating system of quantum machines, the R&D team can manage fragile quantum bits more efficiently. This move has attracted the UK’s National Physical Laboratory (NPL) and multiple international top universities to join the ecosystem, jointly advancing the development of practical quantum applications.

These open models, data, and frameworks are available on GitHub, Hugging Face, and build.nvidia.com.

This article, “NVIDIA Releases the World’s First ‘Ising Model’ Built for Quantum AI,” first appeared on Lian News ABMedia.