Embedding 3 Small
Available nowAn accessible starting point for everyday search, recommendations and organizing information.
What it does
OpenAI’s text-embedding-3-small turns text into a representation of its meaning. Related questions, descriptions and documents can then be compared, making it useful for search, recommendations, clustering and classification.
It is the economical member of OpenAI’s third-generation embedding family. Its default vectors have 1,536 dimensions, and the model supports shorter output vectors when you want to reduce index storage. The official guide describes improved multilingual performance compared with OpenAI’s earlier embedding generation.
Language coverage
OpenAI reports multilingual retrieval results on MIRACL, an 18-language benchmark. Benchmark inclusion is evidence of evaluation, not a guarantee for every dialect or business collection.
Official language coverage ↗Where it helps
Everyday search
Connect a customer question with relevant help or product information.
Recommendations
Suggest items or articles with related descriptions.
Organizing feedback
Group similar customer comments to identify recurring themes.
When to choose it
A straightforward OpenAI option for a first search or recommendation feature. Start with the default output, then evaluate a shorter vector if storage becomes a meaningful cost.
For difficult multilingual retrieval, compare it with Embedding 3 Large using representative queries. An embedding finds related information; it does not write answers or decide whether a statement is true.
Official sources
Paraphrased from the provider’s documentation and original model information. Selection advice is our editorial guidance.
OpenAI · Official release ↗OpenAI · Embeddings guide ↗Reviewed October 3, 2026