Vertex AI Multi-Modal

Overview

Google's Vertex AI platform exposes multimodalembedding@001, a multimodal embedding service that maps text, images, and video into a shared semantic space. It is intended for retrieval and semantic similarity tasks that benefit from comparing content across modalities in a managed API workflow.

Architecture

Google Cloud documentation describes multimodalembedding@001 as a multimodal embedding service and documents its vector dimensionality, but does not provide a detailed public architecture description. The service returns 1,408-dimensional embeddings by default and also supports lower output dimensions.

Capabilities

The embeddings support semantic search, recommendation, content moderation, classification, and similarity-based retrieval across text, image, and video. Text and image embeddings share the same dimensionality and semantic space, enabling cross-modal use cases such as text-to-image retrieval.