New revenueFlagship

Recommendation Systems

The model generates recommendations straight from the catalog. It can only ever answer with things that exist.

Pick a stage

talk trackDescribe a mood, not a keyword. Both models read the same listening history and generate matching tracks side by side, from the same catalog.

Try a query

Pick a query above to retrieve tracks from both models in parallel.

Catalog: 8,165 tracks from the public Million Playlist Dataset (RecSys Challenge 2018) — 7,143 seen in training, 1,022 held out. Same data, same trie, same catalog for every model.