Embedding Dimensions: What 768 vs 1536 Means and the Tradeoffs
~12 min read
The dimension count of an embedding is how many numbers describe each piece of text — more dimensions can capture finer distinctions in meaning, at the cost of more storage and slower search.
Embedding Dimensions: What 768 vs 1536 Means and the Tradeoffs is a Pro topic
Sign in, then upgrade to Pro or Power to unlock this topic and the full AI Engineering curriculum.
Key points
- •The dimension count is how many numbers make up each embedding vector — 768 dimensions means every embedding from that model has exactly 768 numbers
- •No single dimension maps to one human-readable concept — meaning emerges from directions and distances across many dimensions working together
- •Common sizes: 384 (small/fast models), 768 (BERT-base and mid-sized models), 1536/3072 (OpenAI's larger embedding models)
- •More dimensions give more 'room' to separate subtly different meanings, but cost more storage and slower distance computation at scale
- •Choosing dimension count is a real engineering tradeoff between retrieval quality and storage/latency — not simply 'bigger is always better'