AlphaQubit

Published in Nature in December 2024 (DOI 10.1038/s41586-024-08148-8), AlphaQubit is a collaboration between Google DeepMind and Google Quantum AI that applies transformer neural networks to quantum error correction — the central obstacle to useful quantum computers. Trained first on hundreds of millions of simulated error examples and fine-tuned on real Sycamore processor data, it made 6% fewer decoding errors than the best slow tensor-network decoder and 30% fewer than fast correlated matching, scaling tests up to 241 qubits.

AlphaQubit is a research system, not a downloadable product: there are no public weights, and the team notes it is currently too slow for real-time correction in superconducting processors. Its significance is methodological — showing that learning-based decoders can beat classical algorithms on real quantum hardware data — rather than as a deployable tool today.