Gaia Data Release 3. All-sky classification of 12.4 million variable sources into 25 classes

Gaia DR3 contains 1.8 billion sources with G-band photometry, 1.5 of which BP and RP complemented by positions on the sky, parallax, proper motion. The median number field-of-view transits in three photometric bands is between 40 44 measurements per source covers 34 months data collection. We pursue a classification Galactic extra-galactic objects that are detected as variable across whole sky. Supervised machine learning (eXtreme Gradient Boosting Random Forest) was employed to generate multi-class, binary, meta-classifiers classified time series G, BP, bands. Classification results comprise 12.4 million (selected from much larger set potential objects) include about 9 stars into 22 variability types Milky Way nearby galaxies such Magellanic Clouds Andromeda, plus thousands supernova explosions distant galaxies, 1 active galactic nuclei, almost 2.5 galaxies. identification made possible artificial extended Gaia, so they were published galaxy_candidates table archive, separate classifications genuine (in vari_classifier_result table). latter 24 classes or class groups periodic non-periodic variables (pulsating, eclipsing, rotating, eruptive, cataclysmic, stochastic, microlensing), amplitudes few milli-magnitudes several magnitudes.