Catalyst‑Q Get an API key

Run any circuit over the universal gate set. Get the exact probability of every outcome, or a refusal that says why.

Paste OpenQASM 2 or 3. The answer comes back exact, with a public record you can cite. The first run needs no account.

Experimental, for research and evaluation. The limits are published next to the results: read the scope.

Run a circuit, no account See the measured record

Run a circuit here. No account.

A GHZ chain across 40 qubits with T, rotations and a Hadamard on top, measured on four qubits along it.

Any gate in the universal set, measurements at the end. Without an account a circuit may have up to 64 qubits and 4,000 gates, and up to 20 measured qubits. A run that will not fit is refused with the numbers, never approximated; the scope says where that happens. Six runs a minute and sixty a day from one address.

Random circuit, depth 4, mid-chain ⟨Z⟩
1,000 qubits · 0.5 s · exact
TFIM quench, order parameter, 200 sites
7.7 s · exact
Surface code d=9, code-capacity threshold
4 ms per shot · 81 qubits
Width-side record, operator identity
1,000,000 qubits · 3.2 s

Observable-side record cq-front-2026-09-01, measured 2026-09-01 against numpy dense statevector (n ≤ 20), qiskit-aer 0.17.2 matrix-product state (n ≥ 24), Majorana-chart Pfaffian (matchgate circuits). Raw record. Width-side record cq-prod-2026-08-22-ed64918e, revision 5a0d43a69f5e. Both on the evidence page.

Four things you can run here, and what each one returns.

Any circuit, every outcome.

OpenQASM 2 or 3, any gate in the universal set, measured at the end. The exact probability of every outcome on up to 20 measured qubits, in the bit order Qiskit prints, with a public record of the run.

Run one above

Expectation values on deep circuits, with a certificate.

Trotterised dynamics, variational ansätze, random brickwork. Each value comes back with the bound on its own error, 0 when it is exact. Matchgate circuits are recognised and computed exactly. Every job returns a citable record.

Open the route

Coherent errors on the surface code, exactly.

Stabiliser simulators cannot represent a coherent rotation, and dense ones stop near 25 qubits. Here the syndrome-resolved logical channel is exact at d=3 and d=5 (d=7 offline), next to threshold curves at any distance.

Open the lab

Operator identity where no simulator can follow.

Did the transpiler preserve this operator? Is the circuit and its inverse the identity? Answered exactly at a million declared qubits in 3.2 seconds, against analytic truth, in CI.

Read the reference

scope

What it does, and what it does not.

Stating the boundary next to the result is the point. In a field where every benchmark is self-reported, the vendor that publishes its own limits is the one worth checking.

In class

  • Any circuit over the universal gate set, measured at the end: the exact probability of every outcome on up to 20 measured qubits, or up to 10 when the measured qubits depend on more than 20 qubits of the circuit. Counts, when asked for, are those probabilities times the shots, rounded.
  • Exact expectation values and typed observables on structured circuits, at widths of a million qubits.
  • Local observables on deep circuits: exact to depth 10 on a 1D brickwork of any width (measured at 1,000,000 qubits), certified beyond it. The cost is set by depth, not width.
  • Condensed-matter quenches: certified observable transport (TFIM to 100 sites within 4e-7 of an MPS oracle) and exact matchgate dynamics in the fermionic chart (TFIM order parameter at 200 sites in 7.7 s).
  • QEC: code-capacity thresholds at any distance in O(n) per shot, and the exact coherent-error logical channel of the rotated surface code at d=3 and d=5.
  • Analytic ground truth a third party can check without trusting us.

Out of class

  • Random sampling of shots: counts are exact probabilities times the shots, rounded, never draws. Distributions over more than 20 measured qubits are refused.
  • Scrambled circuits wider than 20 qubits at depth of order the width: each measured qubit depends on more of the circuit than one run holds, and the run is refused with the numbers, never approximated.
  • Interacting spin models at long times (XXZ, Hubbard): the value is returned with its certificate, and past 24 sites the certificate stops certifying.
  • Any claim about hardware fidelity. This is a classical engine, not a QPU.

Who uses it.

Condensed-matter and many-body groups

Quench dynamics past the state-vector ceiling, with a number you can put in a paper and a certificate a referee can check.

QEC teams

Coherent-error logical channels that stabiliser tools cannot express; threshold sweeps at any distance in milliseconds per shot.

Algorithm teams with metered hardware

The observable you were about to pay a QPU to estimate, computed first. One prevented $6,000 run pays for a year.

Toolchain vendors and assurance

A correctness oracle in CI above the width your own simulator reaches, and an independent verification record for a claim you are evaluating.

Run your own circuits.

Create an account, open the console, and run your own circuits and observables. The free tier needs no card.

Start free Run one first, no account