Cloudflare Vectorize Now Supports Up to 20 Million Vectors per Index!

Hi, I'm Shiichan! Today I've got a nice capacity upgrade to share for Cloudflare's vector search service, Vectorize!

Cloudflare Changelog developers.cloudflare.com

What was announced?

According to the Cloudflare Changelog, the maximum number of vectors you can store in a single Vectorize index has been raised to 20 million. That's exactly double the previous limit.

The story so far

Previously, a single Vectorize index could hold up to 10 million vectors. If you wanted to run semantic search, recommendations, or retrieval-augmented generation (RAG) over a larger dataset, you sometimes had to split your data across multiple indexes to fit under that ceiling.

What changes

With the new limit, a single index can now hold up to 20 million vectors. That means use cases that used to require splitting data across indexes can now keep everything in one place, cutting down on the management overhead of splitting and merging.

This benefits things like:

• Large-scale semantic search
• Recommendation systems
• Retrieval-augmented generation (RAG) applications

Dive Deep

Vector dimensions and precision haven't changed: Vectorize still supports up to 1,536 dimensions per vector at 32-bit precision. The only thing that changed is how many vectors fit in a single index.

For the full details, check the Vectorize limits documentation before you design a large index.

Wrap-up

• Vectorize's per-index vector limit doubled from 10 million to 20 million
• Dimensions (up to 1,536) and precision (32-bit) stay the same
• Semantic search, recommendations, and RAG apps handling large datasets no longer need to split across multiple indexes

If you've been wrestling with splitting datasets across Vectorize indexes, this update is exactly what you were waiting for!