Classiq + CUDA-Q: Faster Hybrid Quantum Workflows

Quantum algorithm development is highly iterative. Teams move through cycles of modeling, circuit generation, simulation or execution, evaluation, and refinement. Each cycle reveals how an algorithm behaves, how it scales, and how sensitive it is to parameter choices. The speed of this development cycle often determines how quickly new ideas can be explored

Classiq and NVIDIA recently demonstrated an integration between the Classiq platform and NVIDIA CUDA-Q that connects high‑level quantum modeling with hybrid execution workflows. Developers can generate CUDA-Q kernels from Classiq quantum programs and run them within CUDA-Q environments designed for hybrid quantum‑classical workloads. In a recent benchmark, a workflow that previously required about 67 minutes completed in roughly 2.5 minutes on a single NVIDIA A100 GPU.

In this post we’ll look at what the integration enables and why faster iteration matters when developing hybrid quantum algorithms.

The Classiq–CUDA-Q Integration

The Classiq platform focuses on high‑level quantum algorithm design. Developers describe quantum programs using functional models rather than constructing circuits gate by gate. The platform then synthesizes optimized quantum circuits based on the constraints of the target environment

CUDA-Q, developed by NVIDIA, provides a framework for building and running hybrid quantum‑classical programs. It enables quantum kernels to execute alongside classical code while leveraging accelerated computing resources such as GPUs.

The integration connects these two stages of development.

Developers can generate CUDA-Q kernels from Classiq quantum programs and run them within CUDA-Q hybrid execution environments.

This makes it easier to integrate high-level algorithm models created in Classiq with CUDA-Q hybrid workflows.

Why Iteration Matters

Quantum algorithm development relies heavily on experimentation. Many design questions can only be answered by running multiple experiments and comparing results.

Researchers often evaluate questions such as:

• Quantum algorithm development relies heavily on experimentation. Many design questions can only be answered by running multiple experiments and comparing results.
• How does circuit depth change as system size increases?
• Which algorithm variant performs better for a given problem instance?

Hybrid algorithms intensify this requirement. Techniques such as Variational Quantum Eigensolvers, QAOA, and Iterative Quantum Amplitude Estimation involve classical optimization loops that repeatedly invoke quantum circuits.

Each optimization step may require executing or simulating a circuit. In practice, algorithm development can involve hundreds or thousands of circuit evaluations.

When iteration cycles are slow, experimentation slows as well. Faster execution environments allow developers to test more variations, explore parameter spaces more thoroughly, and refine algorithms more efficiently.

Reducing iteration time therefore has a direct impact on how quickly teams can evaluate and improve hybrid quantum algorithms.

CUDA-Q and Hybrid Execution

CUDA-Q was designed to support hybrid quantum‑classical computing within accelerated computing environments. The framework allows developers to write quantum kernels that integrate with classical programs, enabling hybrid workflows where classical and quantum components interact during execution.

One advantage of CUDA-Q is its ability to take advantage of GPU acceleration. GPUs allow many circuit evaluations to run in parallel during simulation, which is particularly useful for hybrid algorithms that repeatedly invoke quantum circuits.

CUDA-Q can also be used within GPU‑accelerated computing environments commonly used for simulation, optimization, and data processing.

By connecting Classiq-generated programs to CUDA-Q workflows, developers can run Classiq models within CUDA-Q hybrid execution environments.

From Classiq Program to CUDA-Q Kernel

The integration enables developers to generate CUDA-Q kernels from Classiq quantum programs.The workflow begins with a high‑level quantum model defined in the Classiq platform. Instead of manually constructing circuits, developers describe the logical structure of an algorithm using Classiq modeling tools. The platform synthesizes circuits that satisfy constraints such as qubit count and compatibility with the target environment while optimizing characteristics such as circuit depth.With the integration, Classiq programs can be used to generate CUDA-Q kernels that run inside CUDA-Q hybrid programs.These kernels can then be combined with classical control logic, optimization routines, or additional CUDA-Q kernels within the same hybrid workflow. This allows teams to integrate Classiq‑generated components into broader hybrid programs built with CUDA-Q.The integration is available through the Classiq development environment, allowing developers to generate CUDA-Q kernels from Classiq programs and incorporate them into CUDA-Q workflows.

Benchmark: IQAE Options Pricing

To evaluate the workflow, the teams ran a benchmark based on Iterative Quantum Amplitude Estimation (IQAE), a technique often used in financial modeling.

The benchmark involved an options‑pricing problem implemented as a 31‑qubit quantum circuit. The circuit was synthesized using the Classiq platform and executed through CUDA-Q using an NVIDIA A100 GPU.

In the demonstration, circuit synthesis and execution previously required about 67 minutes under the prior workflow configuration.

Using the updated integration and CUDA-Q execution environment, the same workflow completed in approximately 2.5 minutes.

The improvement reflects GPU-accelerated simulation through CUDA-Q together with a tighter connection between modeling and execution.

For developers, the most visible effect appears during experimentation. Experiments that previously required significant waiting time can now run much more quickly, allowing more algorithm variants to be evaluated within the same development window.

What Faster Iteration Enables

Shorter iteration cycles allow research teams to explore algorithm behavior more thoroughly.

Developers can run more experiments to compare algorithm variants, evaluate parameter choices, and study scaling behavior as circuit sizes grow. Faster feedback can also help teams identify bottlenecks and refine algorithm designs earlier in the development process.

These capabilities are relevant across many areas where hybrid quantum algorithms are being studied, including financial modeling, optimization, scientific simulation, and algorithm benchmarking.

As hybrid quantum workflows evolve, development environments that support rapid experimentation will help researchers evaluate algorithm ideas more efficiently.

The Classiq and CUDA-Q integration shortens the path between modeling and hybrid execution in GPU‑accelerated computing environments.

Getting Started

The CUDA-Q integration is available through the Classiq platform.

Developers can model quantum programs in Classiq, generate CUDA-Q kernels, and run hybrid workflows using CUDA-Q environments. This enables experimentation with hybrid algorithms, benchmarking workflows, and large‑scale simulation runs.

To learn how to generate CUDA-Q kernels from Classiq programs and run hybrid workflows, see the Classiq documentation for step‑by‑step instructions.

Shorter development cycles allow developers to explore algorithm ideas more quickly. Connecting high-level modeling tools with accelerated execution environments reduces the time required to run hybrid experiments.