Nvidia has released Ising, a family of open-source AI models designed to solve two critical bottlenecks in quantum computing: calibration and real-time error correction. The models integrate with Nvidia's CUDA-Q quantum software platform and are now available on GitHub, Hugging Face, and build.nvidia.com. This move signals Nvidia's strategy to embed itself deeper into the quantum computing ecosystem without building quantum hardware itself .

Why Quantum Error Correction Matters Right Now?

Today's quantum processors are fragile. According to Nvidia, the best quantum processors produce an error roughly once every thousand operations. The challenge isn't just detecting these errors; it's correcting them fast enough to keep the quantum computer functional. Sam Stanwyck, director of quantum product at Nvidia, explained that the logical error rate is directly tied to how quickly decoding runs alongside the hardware .

"The logical error rate is directly tied to how quickly decoding runs alongside the hardware," noted Sam Stanwyck, director of quantum product at Nvidia.

Sam Stanwyck, Director of Quantum Product at Nvidia

This is where Ising comes in. The Ising Decoding family comprises two variants of a 3D convolutional neural network, optimized for speed and accuracy respectively. Nvidia has benchmarked the decoder at 2.5 times faster and three times more accurate than pyMatching, which is the open-source decoder that most quantum research groups currently use. Even more impressive, the Ising decoder requires ten times less training data . A 2.5 times speedup in decoding raises the ceiling on how many gate operations a quantum processor can sustain before its logical qubits break down, making quantum computers more practical.

How Ising Tackles Quantum Computing's Two Main Challenges

• Calibration: Ising Calibration is a 35-billion-parameter vision-language model fine-tuned to read experimental measurements from a quantum processing unit and infer the adjustments needed to tune it. This reduces calibration time from days to hours when paired with an agent, according to Nvidia .
• Decoding: The Ising Decoding variants perform pre-decoding for surface-code quantum error correction, translating redundant measurements from an error-corrected logical qubit into a correction signal that keeps pace with the rate at which new errors appear on the processor .
• Integration: While the models themselves are open-source, they sit atop Nvidia's proprietary stack. The decoder requires NVQLink's low-latency interconnect to feed measurement data to a GPU inside the decoding window, and calibration workflows run through CUDA-Q .

This design reflects Nvidia's broader strategy with models like Nemotron, Cosmos, and GR00T: open the models but keep the surrounding platform proprietary. By doing so, Nvidia drives GPU dependencies throughout the workflow, remaining deeply integrated with the quantum computing industry despite not building quantum hardware itself .

Who's Already Using Ising?

Adoption has been swift among leading quantum research institutions and companies. Named adopters include Fermilab, Harvard, the UK National Physical Laboratory, Lawrence Berkeley National Laboratory's Advanced Quantum Testbed, IQM Quantum Computers, Infleqtion, and IonQ, which is using Ising Calibration directly . This broad adoption suggests that the quantum computing community sees real value in Nvidia's approach to solving calibration and error correction challenges.

The release of Ising represents a significant step forward for practical quantum computing. By dramatically reducing calibration time and improving error correction speed and accuracy, these models could help quantum processors move closer to the fault-tolerant systems that researchers have long pursued. For enterprises and research institutions exploring quantum computing, Ising offers a concrete tool to improve their quantum hardware's performance without waiting for breakthroughs in quantum processor design itself.