NVIDIA Ising: Open-Source AI Family for Quantum Error Correction

NVIDIA has unveiled Ising, the first open-source family of AI models specifically designed for quantum calibration and error correction. This release aims to accelerate the transition from NISQ (Noisy Intermediate-Scale Quantum) devices to truly fault-tolerant quantum computers.

Named after the Ising model in statistical mechanics, these models are optimized to run on NVIDIA Grace Hopper systems, providing real-time feedback loops for superconducting qubits.

Taming Quantum Noise

Quantum computers are notoriously susceptible to environmental noise, which causes decoherence and computational errors. The Ising-7B and Ising-70B models use reinforcement learning to predict and counteract bit-flip and phase-flip errors before they propagate through a circuit.

Key Features:

• Sub-Millisecond Inference: Essential for active error correction cycles in cryogenic environments.
• Hardware Agnostic: Supports IBM, Google, and IonQ quantum architectures.
• Open Weights: Encouraging global research collaboration to solve the most difficult problem in quantum physics.

By open-sourcing Ising, NVIDIA is positioning itself as the foundational software layer for the Quantum-AI stack.