Automatic Scalar and Binary Quantization now supported

Henry_Weller · 2024-12-02T20:46:16.553Z

Happy Monday

We’re excited to announce that three new Atlas Vector search features all related to quantization enter public preview today:

Automatic scalar quantization
Automatic binary quantization with rescoring
Ingestion of BSON binData vectors with subtype int1

These capabilities are all available to newly created clusters, and on a rolling basis will be made available to all existing clusters over the next two weeks. More details on how to leverage them to support massive vector indexes in the 100M or 1B+ range in the revamped quantization docs page here and the announcement blog here.

Please let us know what you think of this feature here, thanks!

Jack_Woehr · 2024-12-04T16:00:38.113Z

The link to the quantumization page is 404

Henry_Weller · 2024-12-12T19:52:02.395Z

Hi @Jack_Woehr! The docs page should be working now

Henry_Weller · 2025-04-29T21:37:22.765Z

Hello!

These three features are now all Generally Available, with the added benefit of being able to view the approximate “Required Memory” of each vector index within the “list search indexes view” of the Atlas UI. This should make sizing when using quantized indexes simpler.

Screenshot 2025-04-23 at 5.44.35 PM1878×932 85.1 KB

Thank you!