Infleqtion said on September 14, 2026, that its researchers had constructed and validated a quantum error-correcting code using an early-access version of NVIDIA’s CUDA-Q Logical software. In a technical blog post, Infleqtion’s Victory Omole wrote that one block of the code stores 18 logical qubits in 98 physical data qubits.
A logical qubit is a unit of quantum information spread across several physical qubits so that errors can be detected or corrected. In Infleqtion’s neutral-atom machines, each physical qubit is a single atom held in place by a laser beam.
Infleqtion has built and compiled the code only in software, counting the 98 physical data qubits but not the extra qubits that would measure its checks on hardware.
Infleqtion made its announcement during IEEE Quantum Week in Toronto. NVIDIA made CUDA-Q Logical available on GitHub the same day and described it as an orchestration layer for building fault-tolerant applications on its open-source CUDA-Q platform. A fault-tolerant quantum computer keeps computing correctly despite errors in its components.
The code
Omole wrote that the team started from a 7-bit cyclic simplex code with [7,3,4] parameters, a classical code for ordinary bits. From it, the team built a hypergraph-product simplex (HGPS) code with parameters [[98,18,4]]. In that notation, 98 is the number of physical data qubits, 18 the number of logical qubits, and 4 the code’s distance. Omole described those parameters as published, citing a preprint by Willers Yang and colleagues.
A code’s distance is the smallest number of physical errors that can change the stored information without being detected. A distance-4 code is guaranteed to detect up to three errors and to correct any single error.
The HGPS code belongs to the quantum low-density parity-check (qLDPC) family, in which each error check involves only a few qubits and one block can store several logical qubits. Omole wrote that the code has 49 X-type and 49 Z-type checks, each involving six physical qubits, and a code rate of about 18.4%. The code rate is the number of logical qubits divided by the number of data qubits.
The team verified in software that the code’s X-type and Z-type checks commute, so they can be measured without disturbing one another. Using CUDA-Q Logical, the researchers then compiled a joint measurement of two logical qubits stored in the same block. “qLDPC software supplies reusable code mathematics, while CUDA-Q Logical supplies the logical ownership, block allocation, and compilation model,” Omole wrote.
What the count includes
Infleqtion said in its release that the code uses “approximately six physical data qubits per logical qubit,” or about 5.4 for the [[98,18,4]] block, and “up to five times fewer physical data qubits than prevailing surface-code approaches.” The company called that “about a fivefold improvement in code rate over the surface-code approach used for comparison” but did not say which surface code it compared against. In the surface code, each patch of physical qubits stores a single logical qubit, as in a recent surface-code experiment on IBM processors.
Syndrome ancillas are extra qubits that measure a code’s checks. Omole wrote that the team left syndrome ancillas out of the 18-in-98 ratio, along with atom movement, scheduling, readout, control, and classical decoding, the computing step that turns check results into corrections. “Those numbers are promising, but they are not a hardware resource estimate,” Omole wrote of the code’s parameters.
CUDA-Q Logical
NVIDIA said Fermilab researchers shortened fault-tolerant algorithm development “from five months to three weeks” with CUDA-Q Logical. NVIDIA also said Iceberg Quantum used the software to model its architecture on Diraq’s silicon-based qubits. In that model, 1,000 logical qubits would require about 150,000 physical qubits, “roughly 10x fewer than” Diraq’s previous estimates.
Infleqtion is integrating its qLDPC library with CUDA-Q Logical “because the code and the machine must be designed together,” said Pranav Gokhale, the company’s chief technology officer and general manager of quantum computing. Sam Stanwyck, NVIDIA’s director of quantum product, said in Infleqtion’s release that the integration “shows how researchers can design and evaluate logical qubit workloads end to end.”
Next steps
Infleqtion said it open-sourced its qLDPC library in 2025, working with researchers from JPMorgan Chase. The company said the HGPS construction follows architecture work for reconfigurable neutral-atom arrays from the laboratory of Fred Chong, its chief scientist for quantum software.
In reconfigurable arrays, atoms can be moved between positions during a computation. Chong is a co-author of the Yang preprint that Omole cited for the code’s parameters.
Infleqtion said it is “now extending the workflow toward physical execution,” adding syndrome extraction, neutral-atom noise modeling, and decoder benchmarking. The company gave no date for running the code on hardware.
Infleqtion said it would share further updates at Quantum World Congress in College Park, Maryland, where Chief Executive Officer Matt Kinsella was scheduled to give a keynote on September 24. In September 2025, the company set a target of 1,000 logical qubits by 2030.
