NVIDIA Ising: Open AI Models Launched to Stabilize Fragile Qubits for Quantum Computing

NVIDIA announced the Ising family of open AI models on April 14, 2026, designed to address the core challenge of qubit stability in quantum computing. Qubits, the basic units of quantum information, are extremely fragile and react to temperature shifts, electromagnetic noise, and hardware imperfections, often causing calculations to collapse. Ising acts as an AI-driven control layer that automates calibration and error correction, two tedious processes essential for maintaining hardware precision. The models leverage vision language technology and neural networks to interpret experimental signals faster than human-led methods, enabling more reliable quantum operations.

Ising introduces two key workflows: Ising Calibration and Ising Decoding. Ising Calibration can shrink tuning workflows from days to hours, while Ising Decoding runs up to 2.5 times faster and three times more accurately than traditional approaches. These improvements are critical because leading quantum processors still make roughly one error per thousand operations, while large-scale applications may require error rates as low as one in a trillion. By automating these processes, Ising helps researchers run deeper circuits, accelerating the path to fault-tolerant quantum systems. The models are available through the open-source CUDA-Q toolkit, allowing developers to adapt them to specific hardware noise fingerprints.