Amazon Nova Multimodal Embeddings is now available in AWS GovCloud (US-West)

Amazon Nova Multimodal Embeddings has launched in AWS GovCloud (US-West), offering a single model to process text, documents, images, video, and audio. This eliminates the need for multiple specialized models, reducing complexity and costs while improving cross-modal retrieval accuracy. The model supports inputs up to 8K tokens and video/audio segments up to 30 seconds, with segmentation for larger files. Developers can use synchronous or asynchronous APIs to optimize for latency-sensitive or high-volume workloads.

The model maps diverse content types into a unified embedding space, enabling applications such as searching video archives with complex queries or finding product images based on customer questions. It also supports financial documentation containing both infographics and text explanations. Multiple output embedding dimensions allow organizations to balance accuracy, performance, storage, and computation costs according to their needs.

Amazon Nova Multimodal Embeddings is designed for agentic RAG and semantic search, providing a unified approach to handling multimodal data. Organizations can integrate the model into their workflows via Amazon Bedrock in AWS GovCloud (US-West), with options for near real-time or batch processing. The model’s capabilities are intended to streamline cross-modal search and retrieval tasks across various industries.

To access Amazon Nova Multimodal Embeddings, developers can visit the Amazon Bedrock console in AWS GovCloud (US-West). Additional details are available in the user guide, which outlines model specifications, API usage, and implementation best practices. The service is now generally available for production use in the specified region.