Google: gemini-embedding-001 API
The first professional text embedding model in the Google Gemini family. Utilizing Matryoshka Representation Learning (MRL), it allows developers to flexibly adjust output dimensions. It excels in multilingual support (100+ languages) and long-context retrieval, serving as a core component for RAG systems and semantic search.
- Context window: 2,048 tokens
- Max output: 3,072 tokens
- Input: text
- Output: text
- Released: 2025-05-20
Frequently Asked Questions
What is the context window of gemini-embedding-001?
gemini-embedding-001 supports a context window of up to 2,048 tokens.