Amazon Nova Multimodal Embeddings: State-of-the-art embedding model for agentic RAG and semantic search

AWS has announced Amazon Nova Multimodal Embeddings, a state-of-the-art embedding model available in Amazon Bedrock that supports crossmodal retrieval across text, documents, images, video, and audio.

• First unified embedding model supporting multiple content types in a single model
• Supports context length up to 8K tokens and text in 200 languages
• Offers four output embedding dimensions (3072, 1024, 384, 256)
• Enables semantic search, similarity comparison, and retrieval-augmented generation (RAG)
• Provides synchronous and asynchronous API options
• Includes content safety filters and bias reduction measures

The model is initially available in the US East (N. Virginia) region and can be used with vector databases like Amazon OpenSearch Service and Amazon S3 Vectors for efficient embedding storage and retrieval.