Amazon Nova
Amazon Nova is Amazon’s family of foundation models for text, image, video, document, speech, embedding, and other generative/multimodal workloads.
AI Practitioner focus
- Access Nova models through Amazon Bedrock rather than provisioning model servers.
- Choose among model variants based on modality, quality/reasoning, latency, context length, and token cost; smaller models generally favor high-volume cost-sensitive use.
- Nova models support use cases such as assistants, document/video understanding, tool-using agents, voice experiences, semantic search, and recommendations.
- Embedding models return vectors for retrieval/similarity rather than natural-language answers.
- Supported customization workflows can use Bedrock and/or Amazon SageMaker AI depending on the model and technique.
- Treat model-family capabilities and exact context limits as version-specific; evaluate the current model in the target Region on representative data.