Skip to main content

Hardware

Run on a real quantum computer

This leaves the simulator behind. Your circuit is transpiled into a device's native gate set and executed on physical qubits — superconducting transmon circuits or trapped ions — with a shared queue in front of it and real noise inside it. The results will not be as clean as Aer's, and that is the entire point.

1 · Circuit

OPENQASM 3.0;
include "stdgates.inc";

qubit[2] q;
bit[2] c;

h q[0];
cx q[0], q[1];

c = measure q;

2 · Device

Looking for devices at https://quantum-lab-api.icybay-dd842fc2.eastus.azurecontainerapps.io…

3 · Shots and cost

Compilation

Two-qubit gates dominate the error on every device here, so this is the one setting that changes how much of your result survives. Level 3 costs compile time, not shots and not money. The preview below and the circuit that is actually submitted both use this level.

Fills idle qubit time with an XY4 echo sequence so a qubit waiting its turn dephases less. Free — it changes the pulse schedule, not the shot count. IBM devices only; other providers ignore it.

Randomises a Pauli frame around each measurement and undoes it classically. A device that mistakes 1 for 0 more often than the reverse shifts your whole histogram in one direction; this turns that systematic shift into ordinary variance. Also free — your shots are split across randomisations, not added to. IBM devices only.

Estimated cost

Pick a device to price 1,024 shots.

4 · What to measure

By default the device returns a histogram of bitstrings — the sampler. Add an observable instead and it measures ⟨ψ|H|ψ⟩ for each one, which is what VQE, QAOA and most of quantum chemistry actually need.

Observables

sampler

Adding an observable switches this submission from the sampler, which returns a bitstring histogram, to the Estimator, which returns <psi|H|psi> for each one. Leave the list empty to sample instead. Each Pauli string needs one letter per qubit, and qubit 0 is the rightmost character — Qiskit's ordering, so IZ is Z on qubit 0.

Start from

A preset replaces the whole list. Hover or focus one for what it measures.

No observables. This job will sample and return a histogram.

These checks are a local copy of the server's, not the server's. A clean editor means nothing failed the rules we know about here; the API re-runs all of them against the parsed circuit and has the final say.

5 · Submit

Your hardware runs

Loading your runs…

Compare two runs

Loading your runs…

One last thing before you spend anything: a local simulation is instant, free and exact. If you want to know what a Bell state's distribution looks like, Aer will tell you in milliseconds. Come here when the question is specifically about the machine — how much decoherence eats your fidelity, what a native gate set does to your depth, how long a real queue actually is. Reproducing textbook results on a QPU costs money and teaches you nothing the simulator would not have.