NVIDIA's Ising: First Open AI Models to Fix Quantum Computers

NVIDIA just dropped Ising, the world's first open AI model family built specifically for quantum computing — and it targets the two biggest pain points holding quantum hardware back.

Ising Calibration is a 35B parameter vision-language model that reads qubit data and automates processor tuning, compressing a process that used to take days down to hours. Ising Decoding is a 3D convolutional neural net (two variants: speed or accuracy) for real-time quantum error correction — running 2.5x faster and 3x more accurate than classical approaches.

Both models are open and available now on GitHub, Hugging Face, and build.nvidia.com. Early adopters include Fermilab, Harvard SEAS, Lawrence Berkeley National Lab, and IQM Quantum Computers.

Why it matters: Calibration and error correction are the unsolved bottlenecks between today's noisy quantum hardware and fault-tolerant, production-ready quantum computers — NVIDIA just made both an AI problem.