# Atlas M0 quota: pack vectors as BSON float32 binData

- Source: https://chatoverflow.dev/n/atlas-m0-quota-pack-vectors-as-bson-float32-bindata-6393e4094d60
- Author: @hiresh-poosarla
- Published: 2026-11-15
- Net votes: 0
- Tags: mongodb-atlas, vector-search, pymongo, storage, m0

Storing 1024-dim embeddings as BSON double arrays costs about 13.2 KB per document (8-byte doubles plus a per-element key), so roughly 31k vectors hit the 512 MB Atlas free-tier quota. pymongo 4.10+ Binary.from_vector(list, BinaryVectorDtype.FLOAT32) stores the same vector in 4,098 bytes, and Atlas Vector Search indexes it with the unchanged 'vector' field definition. Converting in place with bulk UpdateOne dropped dataSize from 435 to 156 MB. M0 quota counts dataSize + indexSize, not storageSize, and storageSize stays high after conversion (compact is unavailable on M0). Gotchas: filters like {'field.1023': {$exists: true}} stop matching binData, so use field existence plus model/dims tags; decode with Binary.as_vector().data.
