Google AI Studio may receive support for two advanced models: Gemini Embedding and Imagen 3. These models have been spotted in the hidden models list, though their public availability on AI Studio is uncertain.
Today we are rolling out an experimental Gemini Embedding model for developers with:
— Logan Kilpatrick (@OfficialLoganK) March 7, 2025
– SOTA performance on MTEB (Multilingual)
- Input context length of (3K --> 8K) tokens
– Output 3K dimensions
– Support for over (50 --> 100) languages
More details in 🧵
Gemini Embedding Model
The Gemini Embedding model, specifically gemini-embedding-exp-03-07, is an experimental text embedding model available via the Gemini API. It offers state-of-the-art performance, supporting up to 8,000 input tokens and outputting vectors of 3,072 dimensions, significantly larger than its predecessor, text-embedding-004. This model excels in multilingual support, covering over 100 languages, and features Matryoshka Representation Learning, allowing for vector truncation to balance accuracy and storage. It is currently in an experimental phase with limited capacity.
Imagen 3 Model
Imagen 3, denoted as imagen-3.0-generate-002, is a cutting-edge image generation model accessible through the Gemini API and Vertex AI in Firebase. It generates high-quality images with better detail and fewer artifacts than previous versions. Imagen 3 supports a wide range of styles and formats, making it versatile for various applications. It includes a non-visible digital watermark to identify AI-generated images.
About AI Studio
Google, the company behind these models, is a leader in AI innovation. The Gemini framework is part of Google's efforts to advance AI capabilities, providing developers with powerful tools for text and image processing. The integration of these models into Google AI Studio would further enhance its capabilities in AI-driven applications.
The potential inclusion of these models in Google AI Studio would significantly enhance its text and image processing capabilities, offering developers more advanced tools for AI applications. This could lead to improved performance in tasks like semantic search, text classification, and image generation, making AI Studio a more competitive platform for developers.
