NVIDIA introduced the launch of Ising, the world’s first open supply household of synthetic intelligence fashions designed particularly to speed up the event of sensible quantum computer systems. NVIDIA Ising takes a serious step towards making quantum techniques extra dependable and scalable by addressing two of the most important bottlenecks: quantum processor calibration and quantum error correction.
Though quantum computing has lengthy promised modern capabilities, scaling techniques for real-world usability stays constrained by fragile qubits and excessive error charges. NVIDIA positions AI as a key management layer that may stabilize and optimize these techniques.
In keeping with NVIDIA founder and CEO Jensen Huang, AI is crucial to creating quantum computing a actuality. He defined that the Ising mannequin turns AI right into a “management aircraft, or working system for quantum machines,” enabling extra dependable and scalable quantum GPU techniques.
The Ising mannequin household introduces two core elements:
Ising Calibration: A 35 billion parameter imaginative and prescient language mannequin (VLM) that may interpret quantum processor measurement knowledge (together with experimental plots) and automate calibration workflows, usually at the side of AI brokers. Duties that beforehand took days can now be accomplished in hours. It outperforms different fashions on a devoted quantum calibration benchmark (QCalEval). Ising decoding: A set of 3D convolutional neural networks (roughly 0.9M and 1.8M parameters) optimized for real-time quantum error correction as a pre-decoder of floor codes. These fashions ship as much as 2.5x sooner efficiency and as much as 3x accuracy (decrease logical error fee) in comparison with conventional approaches akin to pyMatching (when utilized in a hybrid setup).
Collectively, these instruments deal with two of an important bottlenecks in quantum system improvement: accuracy calibration and error mitigation.
NVIDIA Ising is launched below a permissive open supply license (NVIDIA Open Mannequin License), permitting researchers and corporations to coach, fine-tune, and deploy fashions utilizing their very own knowledge whereas sustaining full management of their infrastructure.
These fashions are already being adopted by a broad ecosystem of organizations, together with main analysis establishments, nationwide labs, and quantum know-how corporations, together with Fermi Nationwide Accelerator Laboratory, Lawrence Berkeley Nationwide Laboratory, IonQ, Atom Computing, and Harvard College.
Ising seamlessly integrates into NVIDIA’s intensive quantum computing platform, together with the CUDA-Q software program surroundings and NVQLink {hardware} interconnect. The corporate additionally supplies pre-trained fashions, coaching datasets, documented workflows (“cookbooks”), and NVIDIA NIM microservices to simplify deployment and customization.
This full-stack strategy allows hybrid quantum-classical computing techniques the place GPUs and quantum processors work collectively for real-time management and computation.
