Concept
Embedding Dimension
Number of floats in an embedding vector — typically 256 to 4096 for modern models.
Definition
Embedding dimension trades off representational capacity against memory and query latency. Larger vectors capture finer distinctions but require more storage and slower nearest-neighbour search. Matryoshka embeddings let users truncate to lower dimensions without retraining.
Common use cases
- Storage planning
- Latency tuning
- Embedding selection