China Marks Quantum Breakthrough in Fine-Tuning AI Models

09.04.2025

Chinese scientists have successfully fine-tuned a billion-parameter AI model using the Origin Wukong superconducting quantum computer, marking a global first. This groundbreaking achievement took place at the Anhui Quantum Computing Engineering Research Centre, in collaboration with Origin Quantum and various research institutions. The key aspect of the experiment involved using the quantum computer, equipped with a 72-qubit chip, to retrain a large AI model for specialized tasks by processing domain-specific data.

The fine-tuning process leveraged the quantum computer’s ability to perform hundreds of parallel quantum tasks per batch. As a result, the model’s performance saw significant improvements, including a 76% reduction in parameters and an 8.4% boost in training efficiency. Additionally, the accuracy of the model on a mathematical reasoning task increased from 68% to 82%.

The experiment demonstrates the potential of quantum computing to enhance AI efficiency, particularly in reducing the resources needed for training large models. By integrating quantum capabilities with traditional computing, the research shows how both can work together to improve model performance.

Origin Wukong, operational since January 2024, has already completed over 350,000 quantum tasks in fields like finance, biomedicine, and fluid dynamics. This breakthrough is seen as a major step in addressing the computational challenges associated with large-scale AI models, particularly the issue of ‘computing power anxiety.’

This milestone not only highlights the growing potential of quantum computing but also paves the way for its integration into AI applications in areas such as medical diagnosis and financial risk management. The success of this experiment proves that current quantum hardware can support the complex demands of large AI tasks.