NVIDIA has released Ising, an open-source family of AI models designed to accelerate progress toward practical quantum computers. The models target two of the field’s most stubborn bottlenecks: processor calibration and error correction.
Ising Calibration uses a vision language model to automate the continuous tuning of quantum processors, compressing what previously took days into hours. Ising Decoding, built on a 3D convolutional neural network, handles the real-time decoding required for quantum error correction and benchmarks at up to 2.5x faster and 3x more accurate than pyMatching, the current open-source standard.
The release is positioned as infrastructure, not just tooling. CEO Jensen Huang framed it as AI becoming “the control plane — the operating system of quantum machines,” converting unstable qubits into reliable, scalable systems. The open model approach lets enterprises run inference locally, keeping proprietary hardware data off third-party infrastructure.
Adoption is already broad for a day-one launch. Calibration users include IonQ, Fermi National Accelerator Laboratory, Harvard SEAS, Q-CTRL, and the UK’s National Physical Laboratory. Decoding deployments are underway at Sandia National Laboratories, UC Santa Barbara, University of Chicago, Cornell, and several others.
Ising integrates with NVIDIA’s existing quantum stack — CUDA-Q for hybrid classical-quantum workflows and NVQLink for direct QPU-GPU interconnects. Models are available on GitHub, Hugging Face, and build.nvidia.com, with NVIDIA NIM microservices and training data provided to support hardware-specific fine-tuning.
The quantum computing market is projected to exceed $11 billion by 2030.
Leave a Reply