NVIDIA Ising Enables Fully Automated Quantum Computer Calibration with Enhanced In-Context Learning

NVIDIA Releases Open VLM to Automate Quantum Processor Calibration

NVIDIA has released Ising Calibration 1.5, an open-source vision language model (VLM) designed to interpret diagnostic outputs from quantum processors and automate their tuning. The new model advances AI-driven quantum calibration with improved in-context learning, allowing it to analyze unfamiliar diagnostic results without prior training examples. This release is significant for making agentic calibration workflows more practical for deployment directly within local lab environments, thanks to a smaller model size and a new quantized version.

Technical Specifications and Performance

The 31-billion-parameter model introduces an NVFP4-quantized version for the first time, enabling its deployment on a single consumer GPU or an NVIDIA DGX Spark system. According to NVIDIA, the model was trained on diverse datasets from multiple qubit modalities and evaluated using the QCalEval benchmark, which measures its ability to interpret experimental results and recommend next steps. In these evaluations, Ising Calibration 1.5 demonstrated strong performance, outperforming comparable open models and remaining competitive with leading closed models.

• Model Size: 31B parameters, with a BF16 version that is 11.4% smaller than its predecessor.
• Quantization: An NVFP4 version is available for efficient deployment on constrained hardware.
• Performance: Achieves a 10% average improvement in zero-shot reasoning over the next-best open model and an 86.5% improvement over its predecessor in in-context learning (ICL).
• Benchmark: Performance validated on the open QCalEval benchmark for quantum calibration plot understanding.

Ecosystem Integration and Accessibility

NVIDIA is releasing the entire Ising model family under the open OpenMDW License, with full-parameter and quantized checkpoints available on Hugging Face and through NVIDIA NIM. The release is complemented by a ready-to-use agent blueprint on GitHub, which integrates with the NVIDIA Nemo Agent Toolkit to facilitate the rapid setup of automated quantum calibration workflows. By optimizing performance for hardware like DGX Spark and offering a single-GPU deployment path, NVIDIA lowers the barrier to entry for research institutions and quantum hardware companies seeking to leverage AI for QPU operations.

With the release of Ising Calibration 1.5, NVIDIA is executing a classic ecosystem play: creating highly specialized, open-source AI tools for a critical scientific niche to drive adoption of its entire stack, from consumer GPUs and DGX Spark hardware to its Nemo agent software framework, effectively becoming an indispensable partner in the quantum computing R&D pipeline.