Embedding Models

AI models for text embeddings and semantic search

Embedding · OpenAI

text-embedding-3-large
Available
TokenLab price-15% Discount
Input$0.1105Output$0.00/1M
Official price
Input$0.13Output$0.00/1M
text-embedding-3-small
Available
TokenLab price-15% Discount
Input$0.017Output$0.00/1M
Official price
Input$0.02Output$0.00/1M
text-embedding-ada-002
Available
TokenLab price— Discount
—
Official price
Input$0.10Output$0.00/1M

All model series

1 series

Choosing the right embedding model

The right model balances task fit, output quality, latency, and price.

Selection signals

  • Match the model to the input and output you actually need.
  • Compare models that use the same pricing unit.
  • A small test reveals quality and latency differences that a catalog cannot.

FAQ

What makes a good embedding model?

A strong fit matches the task, quality bar, latency target, and budget. A small evaluation set makes quality and reliability differences easier to see.

Can I switch models later?

Yes. Keep the public model ID and request format explicit, and compare alternatives on the same tasks.