Dimensionality
How many numbers are in each vector — the size of the space the model thinks in.
Think of it like
Describing a person by 2 traits vs 500: more dimensions, more nuance, but harder to picture.
Example
An embedding model might output 1,536-dimensional vectors; each of those slots captures some sliver of meaning.
How it actually works
Higher dimensionality lets a model encode more distinctions, but costs memory and compute and eventually hits diminishing returns. Weird things happen in high dimensions — distances flatten out — which is why similarity math has to be chosen carefully.
For product teams
Bigger embedding dimensions can mean richer matching, but also more storage and slower search.
For engineers
The size d of the vector space; trades representational capacity against compute and the curse of dimensionality.
Related
- Vector — The length of a vector.
- Latent space — Defines the latent space.
Read anything AI without the jargon
Look up any term in plain English, or save terms as you read with the free Chrome extension.
Open DecoderAdd to Chrome