Skip to content
Calixo

Vector Embedding Storage Calculator

Estimate the raw storage size of a set of vector embeddings.

Inputs

e.g. 1536 for OpenAI text-embedding-3-small, 768 for many BERT-based models.

Saved Scenarios

— select 2+ to compare

Estimated Storage (GB)

5.72

Spark says

How it's calculated

Formula

StorageGB=Vectors×Dimensions×Bytes per Value10243Storage_{GB} = \dfrac{Vectors \times Dimensions \times Bytes\ per\ Value}{1024^3}
Bytes\ per\ Value
— Storage precision — float32, float16, or quantized int8

What is the Vector Embedding Storage Calculator?

Vector embeddings store one floating-point number per dimension for every item — this calculator estimates the raw data size before adding index structures (HNSW graphs, IVF clusters, etc.) that most vector databases layer on top.

How to use it

  1. 1 Enter how many vectors (embeddings) you'll store.
  2. 2 Enter the embedding's dimensionality.
  3. 3 Choose the storage precision.

Worked examples

1M vectors, 1536 dimensions, float32

About 5.72 GB of raw vector data.

Try it

Limitations

  • •This is raw vector data only — most vector database indexes add 20-100%+ overhead on top for fast search, so actual disk/memory usage will be higher.

Frequently asked questions

Why would I use float16 or int8 instead of float32?

Lower precision cuts storage and memory roughly in half (float16) or to a quarter (int8), at some cost to search accuracy — a common trade-off at large scale.