AI Search storage-optimized endpoints

An AI Search endpoint type for very large indexes, launched as Mosaic AI Vector Search storage-optimized endpoints. It holds over one billion vectors at dimension 768, against about 320 million on a standard endpoint, indexes 10 to 20 times faster and answers queries in about 250 milliseconds. It does not support continuous sync, columns to sync, High QPS or FedRAMP workspaces, the embedding dimension must be divisible by 16, and each sync rebuilds parts of the index. Dedicated full-text search indexes (Beta) run only on storage-optimized endpoints.

Status and dates are extracted from public sources only. We don't see the private AE-shared roadmap. If something looks wrong, tell us.