Quant²AI – An End-to-End Quantum AI Benchmarking Framework for Both Researchers and Practitioners

Quantum AI and Quantum Machine Learning (QML) are among the most promising and dynamic research fields, leveraging quantum principles to enhance computation. However, the vast variety of QML models makes it difficult to assess their performance and identify the most effective approaches. Standardized benchmarking is lacking, and end users often struggle to determine whether quantum AI—and which specific approach—is suitable for their use cases. Addressing these challenges is essential to evaluate the current state of quantum AI and advance toward quantum utility.

In the Quant²AI project—a collaboration between the DLR Institute for AI Safety and Security, Fraunhofer ITWM, Conet Solutions, data cybernetics, and JoS QUANTUM—we have developed a holistic benchmarking framework for quantum AI pipelines. Running on a high-performance cluster, it supports execution on both simulators and real quantum hardware. Our end-to-end approach evaluates not just QML models but the whole pipeline, including e.g., pre-, and postprocessing. It allows systematic comparisons of quantum AI pipelines, including hyperparameter variations and benchmarks against classical AI. Its modular design enables easy integration of new components, such as alternative data preparation. Standardized and real-world datasets ensure comprehensive testing across various use cases. We provide state-of-the-art evaluation metrics and intuitive visualizations, offering benchmarking as a service to the quantum computing community and industry.

Our framework allows researchers to efficiently test newly developed quantum AI components against state-of-the-art solutions in an end-to-end setting. It also provides end users an intuitive way to integrate their specific use cases into the benchmarking process. Quant²AI delivers a fundamental solution for identifying promising quantum AI applications, selecting the most suitable approaches, and equipping non-experts with a clear and accessible assessment of AI quality.