it says on the pricing document says :

Limits.

Each **p1** pod has enough capacity for 1M vectors with 768 dimensions.

Each **s1** pod has enough capacity for 5M vectors with 768 dimensions.

My question is that does it consider with metadata max size 40k or without ? it seems for 5M vector each with 40k metadata. it is about 200GB disk size for S1. is my calculation correct ? Thank you ,

1 Like

It says here in the documentation that this approximation is only for vectors solely and no metadata. If you add metadata to it, the number of vectors you are able to store would reduce.

Now solely based on the vector size memory calculation we can assume 3 GB of storage space. Which would be pretty low for single instance, but I assume this might not be the restriction for metadata as well since they might have this limit due to having to load all vectors in memory for performing nearest matching search, which would ideally not be the case for metadata. But I did my calculations only based on assuming this 3 GB size to be on the safer side.

Only way to validate this would be to actually insert as many records as you can in the single instance and see where it fails. This is the only way I can think of since they have not mentioned anything about metadata size implications on the amount of vectors you could store in single instance.