NVIDIA releases open-source quantum calibration model Ising Calibration 1.5, supporting single-GPU deployment

PANews July 28 — NVIDIA (英伟达) officially released the open-source vision language model (VLM) Ising Calibration 1.5, which is used to automatically interpret quantum processing unit (QPU) diagnostic results and complete calibration, helping enable fully autonomous calibration of quantum computers. The new version supports enhanced In-Context Learning: it can analyze unknown diagnostic results without needing prior training samples, and can improve reasoning ability by combining relevant experimental samples. In the QCalEval benchmark, its In-Context Learning performance improved by 86.68% compared with the previous generation, and the model size (BF16) was reduced by 11.4%.

In addition, Ising Calibration 1.5 first introduces an NVFP4 quantized version that can be deployed on a single GPU or on NVIDIA DGX Spark, while maintaining performance comparable to some leading closed-source models. NVIDIA also open-sourced the model weights, dataset, benchmark tests, and deployment方案 to support development of quantum computing calibration workflow.