In a significant advancement for quantum computing and artificial intelligence, Chinese quantum computing startup Origin Quantum has successfully completed the fine-tuning of a billion-parameter AI model using its third-generation superconducting quantum computer, Origin Wukong. The breakthrough was reported by Science and Technology Daily on Monday.
A First-of-its-Kind Achievement
Based in Hefei, eastern China, Origin Quantum claims this is the first real-world application of quantum computing in large AI model tasks. The achievement demonstrates that existing quantum hardware can now support advanced AI tasks such as fine-tuning, which was previously the domain of classical high-performance computers.
Chen Zhaoyun, a scientist from the Hefei Comprehensive National Science Center, described the development as a “world-first” that showcases quantum computing’s growing potential in AI applications.
Also read: Nvidia to Launch Quantum Lab in Boston
Efficiency Boost: Fewer Parameters, Better Performance
In a remarkable twist, the experiment revealed that reducing the number of parameters in the AI model by 76% actually improved training performance by 8.4%. This suggests that quantum computing can facilitate lightweight AI models that are more efficient and scalable, particularly useful as industries confront rising computing power constraints.
Fine-tuning is the process of adapting a general-purpose AI model—like DeepSeek—to a specific domain such as healthcare or finance by retraining it with relevant datasets.
Meet the Machine: Origin Wukong
The Origin Wukong, a 72-qubit superconducting quantum computer, is China’s independently developed third-generation quantum system. It is programmable, deliverable, and already in use globally. Since its launch, the machine has supported over 350,000 quantum tasks for more than 23 million users across 139 countries via cloud access.
This milestone could mark a turning point in the convergence of quantum computing and artificial intelligence, offering a new pathway to develop more efficient AI solutions as global demand for computing power continues to surge.
