A new orchestration layer helps researchers model, verify, and optimise quantum systems built around logical qubits.
NVIDIA has expanded its open-source CUDA-Q platform with CUDA-Q Logical, an orchestration layer for developing and evaluating applications designed for fault-tolerant quantum computers. The addition provides researchers with a programmable workflow for coordinating quantum algorithms, error-correction methods, hardware architectures, and other system components.
Fault-tolerant quantum computing relies on logical qubits to reduce the impact of errors occurring in physical qubits. However, designing such systems requires multiple layers of hardware and software to be evaluated together. A change in the algorithm or error-correction approach can alter the physical-qubit count, runtime, and other resources required. CUDA-Q Logical addresses this challenge by allowing researchers to model different system configurations and switch between implementation choices within a common workflow.
The platform is intended for applications where quantum processors could eventually tackle computationally intensive problems in drug discovery, financial modelling, materials development, and scientific research. It can also help quantum hardware developers evaluate how different qubit technologies and error-correction strategies affect the resources required to reach a target number of logical qubits.
The key tech features are:
- Programmable fault-tolerant system orchestration
- Hardware-agnostic quantum workflow modelling
- Verifiable resource estimation
- Support for logical-qubit system exploration
- Open-source development environment
Fermilab has used CUDA-Q Logical to convert fault-tolerant system designs into a repeatable computational workflow. According to NVIDIA, the approach reduced the time required to explore physical qubits, runtimes, and error-correction configurations from about five months to three weeks.
Another use case comes from Iceberg Quantum, which modelled a fault-tolerant architecture for Diraq’s silicon-based qubits. The modelling indicated that 1,000 logical qubits could potentially be implemented using about 150,000 physical qubits, substantially below earlier estimates.
The expanded platform also incorporates QUOPS, an open, hardware-agnostic benchmark developed by Sandia National Laboratories. QUOPS is designed to measure quantum-system progress toward utility-scale applications rather than relying only on metrics such as physical-qubit count, fidelity, or coherence. CUDA-Q Logical is available through the NVIDIA CUDA-Q open-source project on GitHub. The platform can also work alongside NVIDIA’s broader quantum-computing ecosystem, including cuQuantum for quantum simulation and NVQLink for connecting quantum processors with GPU-accelerated computing systems.
Click here for the original announcement.
