gate fidelity
Every quantum gate is supposed to perform one exact operation on your qubits — a clean flip, a precise rotation, a perfect entangling step. In real hardware the gate never lands quite where you aimed: control pulses are slightly off, qubits drift, stray noise creeps in. Gate fidelity is the score for how close the gate you actually got is to the gate you meant to apply. It runs from 0 to 1 and is usually quoted as a percentage, so 99.9% fidelity means that on average only about one part in a thousand of the operation went astray. The leftover, one minus the fidelity, is the gate error rate.
Fidelity matters most because errors pile up. A short program may chain together thousands of gates, and a tiny slip on each one compounds into a result you can no longer trust. This is why two-qubit gates — the entangling ones that do the real work — get watched most closely: they are the hardest to make clean, so their fidelity usually sets the ceiling on what a machine can actually run.
Fidelity is also the gatekeeper for fault tolerance. Quantum error correction can fix mistakes, but only once the physical error rate drops below a certain threshold — for the popular surface code, very roughly the 1% mark, meaning gates need fidelity above about 99%. Today's best gates sit right around that frontier, which is genuinely exciting, but clearing the threshold is just the entry ticket: you still need many noisy physical qubits to protect a single reliable logical one. We are not there at scale yet.
Fidelity and error rate are two views of the same number; small per-gate errors add up fast across a long circuit.
Beating the threshold is necessary but not sufficient — a fidelity just over the line still demands a heavy overhead of physical qubits per logical qubit, which is why large fault-tolerant machines remain a work in progress.