Tiger Data launches pg_vectorsearch: 6 ms at 100M vectors

Tiger Data, the company behind TimescaleDB, released pg_vectorsearch, an open-source Postgres extension with a new PRISM index: 6 ms median latency at 95% recall on 100 million vectors. The alpha has known gaps: filtered queries can return fewer rows than LIMIT, and the dimension cap is 1,968.

• PRISM hits 6 ms at 95% recall on 100M vectors; HNSW build failed on 128 GB RAM
• Index uses 25 GB versus 180 GB for HNSW on a 512 GB machine
• Alpha: filtered queries can return fewer rows than LIMIT
• 1,968-dimension cap, no unlogged tables or expression indexes