PRESS RELEASE — QuEra Computing, the leader in fault-tolerant quantum computing, today highlighted research with Harvard University and Anthropic showing that AI can design quantum circuits that require fewer computational resources. The findings show how advances in algorithm design could expand what future fault-tolerant quantum computers can accomplish with a given hardware capacity.
The research preprint, “Discovering state-of-the-art quantum circuits with large language models” (arXiv:2610.22367), examines eight computational building blocks used in larger quantum algorithms. In seven, AI-designed circuits reduced both gate counts and qubit requirements relative to the corresponding baseline designs.
In the eighth, the search found a new approach that offers additional choices between qubit requirements and gate counts. The research, reflecting an industry-first collaboration of AI and Quantum market leaders, was led by Andi Gu as part of Anthropic‘s STEM Fellows Program, with Alek Kemeny of Anthropic, Susanne Yelin of Harvard, and Mikhail Lukin of Harvard and QuEra.
Useful quantum computing arrives where two curves meet. One is hardware capability, which rises as machines gain more qubits and more reliable operations. The other is the resources useful algorithms require, which fall as those algorithms are made more efficient. Progress on either curve moves the meeting point earlier; progress on both can bring valuable applications within reach sooner.
This research addresses the second curve.
The researchers paired Anthropic‘s Claude with an independent automated evaluator that checked each proposed circuit design for correctness and scored its resource use. Claude used that feedback to explore and improve its designs. The evaluator scored each design on two numbers: its Toffoli-gate count and its qubit count. Both numbers matter.
Toffoli gates can account for a substantial share of the overhead in fault-tolerant computations, while logical qubits occupy valuable hardware capacity. Reducing both can lower the resources needed to execute a circuit, although the resulting runtime and physical hardware requirements depend on the system’s architecture. Selected results include:
The combined metric multiplies the Toffoli-gate count by the peak number of logical qubits, so a 39x improvement in that metric is not a 39x speedup. These results are resource estimates for individual building blocks rather than runs on hardware, and their effect on complete applications will depend on how those components are used.
Reducing what key components require can expand the problems a given system could tackle.
Beyond improving existing designs, Claude helped discover a new reversible approach to modular inversion, an arithmetic operation central to elliptic-curve calculations. The approach reconstructs much of the information normally stored during intermediate steps, creating new options for balancing gates against qubits.
In the complete point-addition benchmark, one design required about 17x fewer gates than the prior design using the fewest qubits, while using about 12% more qubits.
“Reversibility makes many efficient classical methods costly to implement on quantum computers. What is exciting here is that AI, guided by a rigorous evaluator, found new circuit structures that work more efficiently within that constraint. This opens a promising path for discovering algorithms for large-scale, useful quantum computers, which QuEra…
