Global First - China Uses Quantum Computing To Fine Tune LLM Model
Fresh & Hot curated AI happenings in one snack. Never miss a byte 🍔
This snack byte will take approx 4 minutes to consume.
Global firsts, are always interesting. It opens up more possibilities and leads to further innovation.
Chinese researchers have now achieved what many quantum computing experts (myself included) considered a theoretical possibility just years away from practical implementation: using an actual quantum computer to enhance artificial intelligence training in a meaningful way.
The Hefei Breakthrough: Where Quantum Computing and AI Converge
Chinese researchers at the Institute of Artificial Intelligence under the Hefei Comprehensive National Science Centre have accomplished something genuinely groundbreaking: they've used China's Origin Wukong quantum computer to fine-tune a large language model with 1 billion parameters.
This isn't just another incremental step forward—it represents the first real-world integration of quantum computing technology with advanced AI training processes.
The numbers speak volumes about this achievement. Using the 72-qubit Origin Wukong quantum computer, the team achieved an 8.4% improvement in training performance while simultaneously reducing the model's parameters by a remarkable 76%.
Having spent years watching both quantum computing and AI evolve separately, I can tell you that these figures represent more than just statistical improvements—they signal a fundamental shift in what's possible when these technologies combine forces.
As Chen Zhaoyun, one of the researchers involved, explained: "This is the first time a real quantum computer has been used to fine-tune a large language model in a practical setting. It shows that current quantum hardware can begin to support real-world AI training tasks."
Perhaps more impressive are the practical improvements in the AI's capabilities. When trained on mental health conversation data, the quantum-enhanced model made 15% fewer mistakes—a significant improvement in a field where accuracy and sensitivity are paramount. In mathematics problem-solving, its accuracy jumped from 68% to 82%—the equivalent of moving from a struggling student to a solid performer in a single technological leap.
Why This Matters: Fine-Tuning and Quantum Advantage
For those who haven't spent their evenings reading academic papers on machine learning (lucky you!), fine-tuning is a critical process that transforms general-purpose AI models like DeepSeek or Qwen into specialized tools for specific applications—whether that's medical diagnostics, financial analysis, or creative content generation.
Traditional fine-tuning methods rely on arrays of power-hungry GPU servers that consume electricity with the enthusiasm of a teenager discovering refrigerators. The process faces inherent limitations in scalability and energy efficiency that have become increasingly problematic as AI models grow larger and more complex.
This is where quantum computing offers a fundamentally different approach. By leveraging quantum mechanical principles like superposition (the ability of quantum particles to exist in multiple states simultaneously—something I've tried and failed to achieve with my work-life balance) and entanglement (where quantum particles become connected and instantly affect each other regardless of distance), quantum computers can theoretically explore vast combinations of parameters in parallel.
The Hefei team developed a novel approach called "quantum-weighted tensor hybrid parameter fine-tuning" to realize this advantage. In this method, the AI model's weights (the parameters that determine how it processes information) are handled by a network that combines quantum and classical computing techniques. The quantum component identifies complex patterns in the data, while the classical component compresses the model to reduce computational demands.
According to Dou Menghan, vice-president of Origin Quantum, the Wukong chip demonstrated remarkable efficiency, with a single batch of AI training data triggering hundreds of parallel quantum operations—showcasing the chip's ability to handle the intensive computational workloads required for AI training.
To appreciate the significance of this development, we need to understand the accelerating global race in quantum computing over the past decade.
The theoretical foundations for quantum computing were established decades ago by visionaries like Richard Feynman and David Deutsch, but practical implementations have followed a more recent and rapidly accelerating timeline:
2011: D-Wave Systems announced the first commercially available quantum computer, though its specialized "quantum annealing" approach limited its applications.
2017: IBM made a 17-qubit processor available to the public through cloud access.
2019: Google claimed "quantum supremacy" when its 53-qubit Sycamore processor performed a calculation that would allegedly take traditional supercomputers thousands of years. (IBM later disputed this claim, suggesting their classical systems could solve it in days with the right algorithm—a debate that continues to highlight the competitive nature of quantum achievements.)
2020: Chinese researchers at the University of Science and Technology of China demonstrated quantum advantage using Jiuzhang, a photonic quantum computer that they claimed performed calculations 100 trillion times faster than classical supercomputers for specific problems.
2021: IBM unveiled its 127-qubit Eagle processor, while Chinese researchers revealed a 66-qubit programmable superconducting quantum processor named Zuchongzhi.
2022: IBM announced its 433-qubit Osprey processor.
2023: IBM reached the 1,000+ qubit milestone with its Condor processor, while quantum error correction made significant strides.
2024: Origin Wukong's deployment for AI fine-tuning represents one of the first practical applications of quantum computing in a commercially relevant context.
And in the race to the best and first to cross the finish line, the quantum computing landscape has evolved into a complex global competition with several key players:
United States
The U.S. has maintained a strong position through both government initiatives and private sector innovation. The National Quantum Initiative Act of 2018 allocated $1.2 billion to quantum information science over five years. Major technology companies like IBM, Google, Microsoft, and Amazon have made significant investments in quantum computing research and infrastructure.
IBM's quantum roadmap has been particularly ambitious, with clear milestones toward increasingly powerful quantum systems. Meanwhile, startups like Rigetti Computing and IonQ have attracted substantial venture capital and developed their own quantum hardware approaches.
China
China recognized quantum technology as a strategic priority somewhat later than the U.S. but has accelerated its efforts dramatically. The Chinese government has reportedly invested over $15 billion in quantum research and development. The National Laboratory for Quantum Information Sciences in Hefei, with a reported $10 billion in funding, represents one of the world's largest investments in quantum facilities.
Chinese universities, particularly the University of Science and Technology of China, have achieved several quantum milestones. The 2016 launch of the Micius satellite demonstrated China's commitment to quantum communications technology, enabling experiments in quantum key distribution across unprecedented distances.
Europe
The European Union launched the Quantum Flagship program in 2018, a €1 billion, ten-year initiative to develop quantum technologies. Individual European countries have also established significant quantum programs, with the UK committing £2.5 billion to its National Quantum Technologies Programme and Germany allocating €2 billion for quantum research and development.
Companies like Pasqal in France and IQM in Finland have emerged as important players in the European quantum ecosystem.
Other Nations
Japan has committed approximately ¥30 billion ($211 million) to quantum technology research through its Quantum Leap initiative. Australia's Silicon Quantum Computing, Canada's D-Wave Systems, and Israel's growing quantum startups have also made notable contributions to the global quantum landscape.
Origin Wukong: China's Quantum Computing Milestone
Launched in January 2024, Origin Wukong represents a significant achievement in China's quantum computing trajectory. As a third-generation superconducting quantum computer with 72 qubits, it stands among the most advanced programmable and commercially available quantum systems globally.
What's particularly notable is Origin Wukong's level of technological self-sufficiency. According to the company, more than 80% of its hardware and software components are domestically produced within China, with in-house backup systems in place—reflecting China's broader push toward technological independence in strategically important fields.
By February 2024, just one month after its launch, the platform had attracted over 20 million visits from users across 139 countries and regions, completing hundreds of thousands of tasks across diverse industries including biomedicine, fluid dynamics, and finance. This rapid adoption speaks to both the platform's capabilities and the growing global interest in practical quantum applications.
Interestingly, international users from the United States were consistently the most active among overseas visitors, followed by those from Russia, Japan, and Canada. This pattern of usage highlights an asymmetry in international quantum computing access that Guo Guoping, a leading quantum physicist and co-founder of Origin Quantum, addressed directly:
"US quantum computers are not open to China. But, adhering to the notion of scientific exploration without borders, we are willing to open our services to users around the world, including the US, to jointly promote the concept of quantum computing for the benefit of mankind."
This statement reveals a fascinating dynamic in the international quantum landscape: while geopolitical tensions have led to restrictions on technology sharing in one direction, scientific collaboration continues to flow in the other. It's rather like refusing to share your umbrella with someone while they happily offer you shelter under theirs—a situation that underscores how scientific progress often transcends political boundaries, even when policies attempt to constrain it.
Also, the announcement that Origin Quantum's fourth-generation quantum computer, Origin Wukong 2, has entered its final development phase suggests that quantum computing capabilities will continue to advance rapidly. This raises important questions about where the integration of quantum computing and AI might lead.
Near-Term Possibilities
In the immediate future, quantum-enhanced AI training could address some of the most pressing challenges in artificial intelligence development:
Energy Efficiency: With AI training increasingly criticized for its enormous carbon footprint (training a single large language model can generate as much CO₂ as several cars do in their lifetimes), quantum approaches that reduce parameter counts and improve training efficiency could significantly reduce AI's environmental impact.
Specialized AI: The improvements in mental health conversation and mathematics problem-solving demonstrated by the Hefei team suggest that quantum computing could be particularly valuable for developing highly specialized AI systems for domains where traditional training methods struggle to achieve desired performance levels.
Smaller, More Efficient Models: The 76% reduction in parameters achieved through quantum-weighted tensor hybrid parameter fine-tuning points toward the possibility of developing more compact AI models that maintain or even exceed the capabilities of their larger counterparts—making advanced AI more accessible for deployment on devices with limited computational resources.
Longer-Term Transformations
Looking further ahead, the quantum-AI convergence could enable more fundamental breakthroughs:
Quantum Machine Learning Algorithms: Beyond enhancing classical AI approaches, entirely new machine learning paradigms designed specifically for quantum computers could emerge, potentially solving classes of problems that remain intractable for classical systems.
Quantum Neural Networks: Theoretical work on quantum neural networks suggests they could exponentially reduce the number of parameters needed to represent complex functions, potentially leading to dramatically more powerful AI systems that require far fewer resources to train and operate.
Simulation of Complex Systems: Combined quantum-AI systems might excel at modeling complex phenomena in chemistry, materials science, and biology—potentially accelerating drug discovery, catalyst design, and our understanding of fundamental biological processes.
As someone who has followed the quantum computing field for years, what strikes me most about the Hefei breakthrough is how it exemplifies both the competitive and collaborative aspects of scientific advancement.
The quantum computing race between nations and corporations has elements of a technological cold war, with enormous investments, strategic secrecy, and clear national security implications. Yet simultaneously, we see researchers sharing their findings and making technologies available across borders, recognizing that scientific progress ultimately benefits from openness and collaboration.
This duality reminds me of the early days of classical computing, when competition drove rapid innovation while an open scientific ethos ensured that advances spread globally. The difference today is the speed at which developments occur and their potential implications for everything from cybersecurity to economic competitiveness.
As Origin Wukong 2 approaches completion and other quantum systems continue to advance worldwide, we stand at an inflection point where quantum computing is transitioning from primarily research interest to practical technology with real-world applications. The integration with AI training represents just the beginning of what promises to be a fascinating new chapter in computational history.
For those of us watching this field evolve, it's rather like witnessing the first controlled flight at Kitty Hawk while simultaneously envisioning transcontinental air travel—we can see both the immediate achievement and glimpse its transformative potential.
The quantum future isn't just coming; in labs from Hefei to Mountain View, from Tokyo to Zurich, it's already arrived. And if this first encounter between Chinese AI and quantum computing is any indication, that future will be both more powerful and more efficient than many anticipated. The race to harness this potential has only just begun, but already the results are, well, quite entangled with possibilities.
About the author: Rupesh Bhambwani is a technology enthusiast specializing in the broad technology industry dynamics and international technology policy. When not obsessing over quantum computing, nanometer-scale transistors and staring at the stars, he can be found trying to explain to his relatives why their smartphones are actually miracles of modern engineering, usually to limited success.