Google’s text embedding model "gemini-embedding-001" is now generally available via the Gemini API and Vertex AI. It costs $0.15 per one million input tokens.
The model supports over 100 languages, handles inputs of up to 2048 tokens, and uses Matryoshka Representation Learning (MRL) to reduce output size, which helps cut memory use and computing costs. Google says the model performs better than its earlier models and external alternatives in several tasks. Since its experimental launch in March, Google states the model has held a top position on the MTEB Multilingual Leaderboard.
