The error threshold (also called the fault-tolerance threshold) is the critical physical error rate below which quantum error correction codes can suppress logical errors — meaning that adding more physical qubits to the code actually reduces the logical error rate rather than adding more noise. Above the threshold, larger codes perform worse because the error correction machinery introduces more errors than it fixes. Below the threshold, exponential error suppression becomes possible.
The threshold value depends on the specific QEC code and the noise model. The surface code has an error threshold of approximately 1% per physical gate under depolarizing noise, one of the highest thresholds among known codes and a major reason for its popularity. Other codes have different thresholds: color codes (approximately 0.1-0.5%), concatenated codes (approximately 0.01-0.1%), and LDPC codes (varying, but potentially competitive with surface codes with lower qubit overhead).
Reaching below-threshold performance has been the central hardware challenge in quantum computing. Physical two-qubit gate error rates have steadily improved across platforms: from roughly 5% in the mid-2010s to 0.1-0.5% today for leading systems. Google's Willow chip (2024) was the first superconducting processor to demonstrate exponential error suppression with increasing code distance, confirming below-threshold operation. This milestone validated the theoretical framework of quantum error correction in practice and established a concrete path to fault-tolerant quantum computing.