Introduction

Introduction

The problem

Useful quantum workloads are rarely one circuit. A variational algorithm evaluates thousands of circuits as it optimises its parameters. A large optimisation problem is split into sub-problems, each solved separately. A chemistry study sweeps over geometries. Every one of those circuits has to be generated, sent somewhere to run, and its result brought back and combined.

The places a circuit can run differ widely. CPU and GPU simulators are fast, exact, and available on demand, but each simulation method has its own limits on size and entanglement. QPUs come from different vendors, each with its own SDK, gate set, calibration data, and queue. Choosing where each circuit should run, and moving circuits there and back in bulk, is work that has nothing to do with the science, and today it mostly falls on the user.

Qoro’s stack takes over that work.

The stack

Layer Component What it does
Programming Divi Open-source Python SDK. Turns a problem (a graph, a QUBO, a molecule, a Hamiltonian, a custom circuit) into batches of circuits, runs them on a backend, and post-processes the results, including the classical optimisation loop.
Scheduling and execution Composer Qoro’s cloud execution layer. Accepts jobs, decides how to lay them out on the target (circuit packing, batching, simulation method), dispatches them to simulators or QPUs, and collects the results.
Simulation Maestro Qoro’s quantum simulator. A C++ engine with CPU and GPU backends, nine simulation methods, and noise models. Runs inside Divi on a laptop, as Qoro’s cloud simulator fleet, or on HPC systems.
Hardware access QPU Interfaces One worker per hardware vendor. Each compiles circuits for its device, submits them through the vendor’s SDK, and returns counts.

Users manage their account, API tokens, credits, jobs, and any QPU systems they have registered through the Qoro dashboard at dash.qoroquantum.net.

Where it runs

The components are the same everywhere. What changes is where the compute runs and who operates it:

Deployment What you get
Open source Divi plus Maestro’s CPU engine on your own machine. No account needed.
Solo Qoro’s multi-tenant cloud, open to anyone who signs up. Divi submits to Composer, which runs jobs on Qoro’s CPU or GPU simulators, or on IBM, IQM, or Braket QPUs using your own vendor credentials. Every account gets GPU and QPU access, with no licence needed.
Maestro Enterprise The Maestro engine with its GPU plugin, installed on your own cluster and run through your own scheduler (for example Slurm).
Dedicato The full stack (Composer, Maestro fleet, QPU access, dashboards) in a single-tenant VPC that Qoro operates, in Qoro’s cloud or in your own.

Deployment Options describes each one technically.

Divi, Maestro’s CPU engine, and Maestro’s Qiskit and PennyLane plugins are open source under the QoroQuantum GitHub organisation.