For AI agents: a documentation index is available at https://www.mongodb.com/docs/llms.txt — markdown versions of all pages are available by appending .md to any URL path.
Docs Menu

MongoDB Vector Search Compatibility & Limitations

This page describes the compatibility and limitations of MongoDB Vector Search features on Atlas. For details on MongoDB Vector Search compatibility with self-managed deployments, see the Deployment Options documentation.

MongoDB Vector Search Feature
MongoDB Version for Feature

8.0+

Query Views directly with $vectorSearch

8.1+

Views with sharded sub-pipelines via $lookup/$unionWith

8.2+

$rankFusion on Views

8.2+

7.0.2+

8.3+

To learn about supported clients, see Supported Clients.

A high index count generates significant load on the base cluster and might disrupt your workload. The hard limit for search indexes on a single cluster is 2,500. However, the number of indexes your cluster can support depends on your cluster tier and workload. Smaller cluster tiers like M10 might experience performance degradation or out-of-memory errors with far fewer indexes. Start with a small number of indexes and monitor your cluster's resource usage as you scale.

You cannot create more than:

  • 3 indexes (regardless of the type, search or vector) on Free clusters.

  • 10 indexes on Flex clusters.

To learn more about index limitations, see:

MongoDB Vector Search indexes support vector embeddings that are less than or equal to 8192 dimensions in length. To learn more about indexing field types, see:

If you're using a clustered collection and have the notablescan parameter set to true, your MongoDB Vector Search indexes may not finish building. To resolve this issue, you must set the notablescan parameter to false or check your log for index status transitions.

Binary quantization is currently not supported for nested vector indexes. Use scalar quantization instead or don't specify any quantization type.

To learn about query option compatibility and limitations, see: