The devices ECCentric compiles codes onto: where their qubits sit, which pairs are coupled and how, and the noise they are simulated with. Pick a device below. Hover over a qubit to see its couplers.
Loading devices…
How the devices are modelled. Every two-qubit gate of a compiled circuit acts on a coupled pair; others need SWAPs, which the compiler inserts. Gates on remote couplers between QPUs, and on long-range couplers of the bicycle module where they are marked as noisy, get the remote error and duration instead of the local ones. Every qubit decays by its T1 and T2 for as long as it waits in each layer of the circuit, for example while other qubits are measured (1.56 µs on IBM devices, as on Heron today; the tenfold-lower IBM model also has threefold longer T1 and T2). On shuttling devices (trapped ions, neutral atoms) any two qubits interact: neighbours in place, others after a move, which costs the move error, its duration and, for the Harvard array, atom loss. The T1 and T2 of the qubits of IBM's planned devices are drawn at random per qubit (T1 around 190 µs, T2 around 130 µs). The layouts follow the IBM roadmap, Bravyi et al. (2024) for the bicycle module, Google (2025) for Willow and Bluvstein et al. (2025) for the Harvard/MIT array.