NVIDIA Ising
NVIDIA Ising is a model family, training framework, and cookbook for building and deploying AI for quantum computing. Leveraging NVIDIA Ising enables quantum computing experts to get the full value of state-of-the-art artificial intelligence technology without requiring machine learning domain expertise. It solves existential blockers and scales useful quantum computing with the best classical computing has to offer.
NVIDIA Ising Models
The NVIDIA Ising model family gives quantum computer builders, operators, and developers the AI tools required to scale devices to fault tolerance. Key challenges on that path include calibration and quantum error correction.
The new Ising vision language model for quantum calibration can be leveraged in an agentic workflow to automate quantum processor bring-up and retune calibration workflows.
Quantum error correction decoders need to be low latency while improving the logical error rate (LER) of the quantum processor they are connected to. Until now, machine learning pre-decoders capable of simultaneously reducing end-to-end latency, improving the LER, and scaling across both space and time to enable efficient lattice surgery operations have not been available. 3D CNN model architectures are key to supporting this capability, and NVIDIA Ising includes a fast, accurate model for SI1000 depolarizing noise models.
Ising Calibration 1
Calibration 1 is 3.27% better than Gemini 3.1 Pro, 9.68% better than Claude Opus 4.6, and 14.5% better than GPT 5.4.
Ising Decoder SurfaceCode 1 Fast
Decoder SurfaceCode 1 Fast offers 2.5x faster latency and 1.1x higher accuracy than PyMatching for d=13, p=0.003.
Ising Decoder SurfaceCode 1 Accurate
Decoder SurfaceCode 1 Fast offers 2.3x faster latency and 1.5x higher accuracy than PyMatching for d=13, p=0.003.
NVIDIA Ising Calibration and Ising Decoding
NVIDIA Ising Decoding includes a training framework for decoder models so you can train your own decoders and custom tailor them to their quantum computer noise models for the best performance. Examples are included for real-time decoding with the models, along with scripts to quantize.
NVIDIA Ising Calibration offers an agentic workflow to enable out-of-the-box ease of use with Ising Calibration 1. Learn how to deploy the model with an agent or integrate with your QPU. Use examples for fine-tuning the models with your own data and the QCalEval benchmark and data to test model effectiveness.
Ising Calibration
Run the Ising Calibration 1 in an agentic workflow with our example on GitHub.
Evaluate your own models with our QCalEval benchmark dataset.
Ising Decoding
Leverage cuQuantum cuStabilizer and PyTorch to efficiently train state-of-the-art models for your specific QPU data, wherever you need.
Get Started With NVIDIA Ising
Start deploying AI models for quantum computing workloads. Explore real-time inference and agentic workflows with real quantum computer data.
Run a Quantum Calibration Agent
Learn how to deploy a quantum calibration agent, input your experiment workflow, and let the agent manage the VLM to evaluate experimental results.
Train Real-Time Decoders
In this self-paced tutorial, learn how to train a pre-decoder specific to your noise model, designed for real-time decoding workflows.
Review the Tutorial
More Resources
Ethical AI
NVIDIA believes Trustworthy AI is a shared responsibility and we have established policies and practices to enable development for a wide array of AI applications. When downloaded or used in accordance with our terms of service, developers should work with their supporting model team to ensure this model meets requirements for the relevant industry and use case and addresses unforeseen product misuse. Please report security vulnerabilities or NVIDIA AI Concerns here.
Get started with NVIDIA Ising today.