Weaviate Cloud Pricing: Vector Dimensions, Storage, Flex and Premium (1 October 2026)

How the Dimension Charge Scales

The dimension meter counts every dimension of every stored vector. One million objects embedded at 1,536 dimensions is 1,536 million dimensions:

This is our arithmetic on the published list floor. Weaviate's table does not state a billing period for the dimension rate; its FAQ says the monthly minimum includes dimension and storage charges, so a corpus this size sits inside the $45 Flex minimum. Because the meter is linear in dimensions, choosing a shorter embedding or truncating a Matryoshka model halves this line when you halve the dimensions.

The September 2026 Price Book Change

Weaviate updated its price book for usage from 1 September 2026. Two things changed: RQ-4 compression became available on HNSW collections with its own rate, sitting between RQ-1 and RQ-8, and backups moved to a corrected per-GiB rate that starts from the same base on AWS, GCP and Azure before regional adjustment. Backups are on by default for every cluster, so the backup line is where existing customers will see the change.

Weaviate vs Pinecone vs Qdrant vs Self-Hosting

Weaviate charges for what you store, not for how often you query it, which suits read-heavy search. Pinecone charges per read and write on top of storage. Qdrant charges for the cluster you provision. Weaviate's hybrid keyword and vector search is included on every plan, including Free.

Weaviate Cloud runs the open-source Weaviate project as a managed service, so the same engine can be self-hosted. The self-hosting break-even and the storage comparison show where running it yourself starts to pay.

Weaviate Embeddings

Weaviate also sells hosted embedding models billed per token, for example Snowflake arctic-embed-m-v1.5 at $0.025 per 1M tokens and arctic-embed-m-v2.0 at $0.040 per 1M tokens. For the cost of generating vectors with other providers, see the embedding model comparison.