NVIDIA announces 'NVIDIA Ising,' an AI for developing quantum computers.

NVIDIA has announced NVIDIA Ising , the world's first open-source quantum AI model family designed to help researchers and companies build quantum processors that can run practical applications.

NVIDIA Launches Ising, the World's First Open AI Models to Accelerate the Path to Useful Quantum Computers | NVIDIA Newsroom
https://nvidianews.nvidia.com/news/nvidia-launches-ising-the-worlds-first-open-ai-models-to-accelerate-the-path-to-useful-quantum-computers

Open AI Models for Quantum Computing | NVIDIA Ising
https://www.nvidia.com/en-us/solutions/quantum-computing/ising/

Introducing NVIDIA Ising, the world's first open AI models to accelerate the path to useful quantum computers.

Researchers and enterprises can now use AI-powered workflows for scalable, high-performance quantum systems with quantum processor calibration capabilities and quantum… pic.twitter.com/jiwig1qsOX

— NVIDIA Newsroom (@nvidianewsroom) April 14, 2026

NVIDIA Ising is an initiative to make open AI models, training frameworks, datasets, and workflows for quantum GPU supercomputing available on NVIDIA's quantum platform. In the field of quantum computing, there has been a lack of dedicated AI models to handle critical processes such as tuning quantum processors and quantum error correction, and NVIDIA Ising is designed to fill that gap and be utilized across the entire quantum ecosystem.

NVIDIA Ising's model family uses AI to accelerate two critical processes: quantum processor calibration and quantum error correction.

Ising Calibration is an open 35 billion-parameter visual language model tuned to infer calibration operations from experimental data on QPUs (quantum processors). It outperformed other methods in all six evaluation tests and, when combined with an AI agent, supports automated QPU calibration. The press release states that automating continuous calibration can reduce the time required from days to hours.

Another Ising Decoding is a set of two open-source 3D CNN models designed for real-time decoding required for quantum error correction. Speed-focused and accuracy-focused versions are available, claiming to be up to 2.5 times faster and up to 3 times more accurate than the current open-source industry standard, pyMatching. Furthermore, in addition to pre-trained models for depolarization noise models for surface coding, a new training framework is provided that can handle any noise model via PyTorch and CUDA-Q.

NVIDIA Ising is released under a generous license, and its data sources, training methods, datasets, and tools for fine-tuning and quantization are all documented. The models can run in researchers' local environments, and can be retrained and tuned for proprietary hardware while protecting sensitive data. Furthermore, a suite of workflows for NVIDIA NIM microservices and quantum computing is provided, making it relatively easy to deploy with minimal preparation.

Furthermore, a new training framework is provided that can handle any noise model, supported through PyTorch and NVIDIA CUDA-Q. NVIDIA Ising is released under a permissive license, with detailed documentation of data sources, training methods, datasets, and tools for model fine-tuning and quantization. This allows developers to tune and train their models themselves using their own hardware and sensitive data.

These AI tools complement NVIDIA's CUDA-Q software platform for hybrid quantum classical computing and NVQLink, which connects QPUs and GPUs. By combining these, NVIDIA aims to lay the foundation for developing today's fragile qubits into future high-speed quantum supercomputers.