Results published: Claude Mythos 5 improved 7 of 8 quantum circuit tasks, with a 17x gate reduction on the elliptic-curve benchmark.
Claude Mythos 5 Cuts Quantum Circuit Gates 17x Across Eight Tasks
More efficient quantum circuits reduce the resources quantum algorithms require. Anthropic says the findings suggest AI can help search for more efficient circuits, though they describe the results as task-specific.
The full picture
Claude Mythos 5 improved resource efficiency across eight quantum circuit design benchmark tasks, reducing both qubit count and Toffoli gate count on seven of them compared to best published designs; on the eighth, it produced a new reversible architecture with better overall tradeoffs. The headline result is a 17-fold reduction in gate count on an elliptic-curve circuit, at the cost of 12% more qubits. Claude iterated on designs using an automated propose-and-test loop optimizing a combined metric of qubit count multiplied by gate count, not measured runtime. The evaluation was run through Anthropic's STEM Fellows Program by Harvard physicist Andi Gu, who gave Claude plain-language instructions. Claude's strongest result applied a classical math shortcut to quantum hardware, supported by new identities Claude proved and verified with formal proofs. Anthropic characterizes the findings as task-specific results suggesting AI can aid circuit search, not as proof that useful quantum computers are now within reach. Separately, Anthropic has open applications for STEM Fellows and Conceptual Reasoning Fellows, four-month positions starting January 2027; the application deadline is October 28.
How it developed
Anthropic opens applications for STEM Fellows and Conceptual Reasoning Fellows, deadline October 28, positions start January 2027.
Sources
Want this in your inbox?
I send one email each morning with the stories that moved. If you would rather just read here, that works too.
Subscribe free