Back to OpenAI
OpenAI /

AI
text-embedding-3-small

API OnlyEmbeddingtextUpdated January 25, 2024

text-embedding-3-small: OpenAI's Efficient Embedding Model

Model Overview

text-embedding-3-small is OpenAI's efficient text embedding model, released in January 2024 alongside text-embedding-3-large. Designed as the cost-optimized alternative for applications where speed and economy are priorities, it produces 1536-dimensional embeddings with significantly improved performance over the legacy Ada-002 model at a fraction of the cost. Like its larger sibling, it supports flexible dimension reduction without significant accuracy loss.


✨ Key Features

Specification Table
FeatureDescription
1536 DimensionsCompact but expressive embedding space
Flexible DimensionsSupports shortening to any dimension ≤1536 via dimensions parameter
Cost-Optimized5x cheaper than Ada-002 with better performance
MTEB Improvement44.0% on MTEB vs. Ada-002's 31.4%
Fast InferenceLower latency than text-embedding-3-large

📊 Benchmarks

Specification Table
Benchmarktext-embedding-3-smalltext-embedding-ada-002
MTEB Overall44.0%31.4%

🔗 Resources

Specification Table

📜 License & Access

Proprietary — Available via OpenAI API.

You might also want to compare

Verified Sources

Model Specs

api-only

Parameters

Unknown

Context Window

Unknown

License

Proprietary

Deployment

api-only

Resources & Links

Curator Notes

Bulk imported from OpenAI developer docs.

Compare Specs

Compare parameters, context windows, modalities, and benchmark scores of this model side-by-side with others.

Compare Model