After we first launched Gemini Embedding 2, we invited builders and enterprises to allow deeper intelligence for his or her initiatives utilizing natively multimodal embeddings. Throughout its preview section, we noticed customers create exceptional prototypes, from superior e-commerce discovery engines to environment friendly video evaluation instruments. These initiatives demonstrated the necessity for methods that may search and cause throughout textual content, picture, video and audio knowledge, which beforehand required advanced, fragmented pipelines.
Now, we’re providing the steadiness and optimizations required to maneuver these multimodal initiatives into manufacturing with the final availability of Gemini Embedding 2 by way of the Gemini API and Vertex AI.
As a core expertise powering many Google merchandise, we’re excited to share these analysis breakthroughs with the developer group.
