CUDA-Q Logical is now available in NVIDIA’s open-source CUDA-Q platform, giving researchers a programmable layer to design, test and verify applications built around logical qubits.
Fermilab used the software to cut fault-tolerant architecture development from five months to three weeks, showing how the layer can speed the codesign of algorithms, error correction and hardware.
Iceberg Quantum modeled Diraq’s architecture with CUDA-Q Logical and estimated 1,000 logical qubits could be built with 150,000 physical qubits—about 10x fewer than earlier estimates.
Sandia National Laboratories also added its QUOPS benchmark to CUDA-Q, offering a hardware-agnostic way to measure how close quantum systems are to utility-scale, fault-tolerant applications.
The release extends NVIDIA’s broader quantum-GPU push, as labs and QPU companies adopt CUDA-Q, NVQLink and related tools to link quantum processors with GPU supercomputing.