Gemini Embedding now generally available in the Gemini API

The Gemini Embedding text model is now generally available in the Gemini API and Vertex AI. This versatile model has consistently ranked #1 on the MTEB Multilingual leaderboard since its experimental launch in March, supports over 100 languages, has a 2048 maximum input token length, and is priced at $0.15 per 1M input tokens.

Google logo
Google
•
Min Choi, Janie Zhang
•3 min read•intermediate•
--
•View Original
EmbeddingGeminiVertex AI

Overview

The article announces the general availability of the Gemini Embedding text model, gemini-embedding-001, in the Gemini API and Vertex AI. It highlights the model's performance on the Massive Text Embedding Benchmark (MTEB) and its versatility across various domains, including science, legal, finance, and coding.

What You'll Learn

1

How to utilize the gemini-embedding-001 model in your applications

2

Why the Gemini Embedding model is superior to previous models and external offerings

3

How to optimize output dimensions using Matryoshka Representation Learning

Prerequisites & Requirements

  • Understanding of text embedding models and their applications
  • Access to the Gemini API and Vertex AI

Key Questions Answered

What is the gemini-embedding-001 model and its capabilities?
The gemini-embedding-001 model is a versatile text embedding model that supports over 100 languages and has a maximum input token length of 2048. It utilizes the Matryoshka Representation Learning technique, allowing developers to optimize output dimensions for performance and storage costs.
How does the gemini-embedding-001 model compare to other models?
The gemini-embedding-001 model has consistently ranked highly on the MTEB Multilingual leaderboard, outperforming both previous Google models and external offerings in various tasks such as retrieval and classification.
What are the pricing and rate limits for using the Gemini API?
The Gemini API offers free and paid tiers, allowing developers to experiment with gemini-embedding-001 at no cost. The pricing for the model is set at $0.15 per 1M input tokens.

Key Statistics & Figures

Maximum input token length
2048
This is the maximum length of input that the gemini-embedding-001 model can process.
Pricing
$0.15 per 1M input tokens
This is the cost associated with using the gemini-embedding-001 model in the Gemini API.

Technologies & Tools

Some links below are affiliate links. We may earn a commission if you make a purchase.

API
Gemini API
Used to access the gemini-embedding-001 model for embedding text.
AI/ML Platform
Vertex AI
Provides a platform for deploying and managing AI models, including gemini-embedding-001.

Key Actionable Insights

1
Leverage the gemini-embedding-001 model for diverse applications in science, legal, finance, and coding.
This model's versatility allows developers to enhance their applications across various domains, making it a valuable tool for improving functionality and user experience.
2
Utilize the Matryoshka Representation Learning technique to optimize your model's performance.
By adjusting output dimensions, you can tailor the model's performance and storage costs to fit specific project requirements, ensuring efficient resource use.

Common Pitfalls

1
Failing to optimize output dimensions can lead to increased costs and reduced performance.
Developers should carefully consider the output dimensions they choose, as using larger dimensions than necessary can waste resources and slow down processing times.

Related Concepts

Text Embedding Models
Machine Learning Applications
API Integration Techniques