← Back to glossary
+Suggest a term
Concept·AI Models & Capabilities·Added 1 day ago

Amazon Nova

Also known as: Nova, Amazon Nova models, AWS Nova

Amazon's first-party model family, available exclusively through Amazon Bedrock. The Nova lineup spans text-only (Micro), multimodal text and image (Lite, Pro), image generation (Canvas), video generation (Reel), and speech (Sonic), plus the extended-thinking Nova 2 generation announced at AWS re:Invent 2025.

Amazon Nova is Amazon Web Services' own foundation model family, built internally and available through Amazon Bedrock rather than competing directly with consumer AI products. Amazon had previously offered Titan models through Bedrock; Nova replaced and expanded that lineup when it launched at AWS re:Invent 2024. The Nova family is designed as the price-performance default for teams already operating on AWS infrastructure, with models fine-tuned for enterprise integration patterns like RAG, agent tools, and structured data tasks.

The Nova lineup as of mid-2026 includes: Nova Micro (text-only, lowest cost), Nova Lite (low-cost multimodal), Nova Pro (multimodal with stronger reasoning), Nova Premier (highest capability), Nova Canvas (image generation), Nova Reel (video generation), and Nova Sonic (speech-to-speech). Nova 2, announced at re:Invent 2025, adds extended thinking capabilities with configurable reasoning levels — Nova 2 Lite supports step-by-step reasoning and task decomposition with a 1 million token context window.

For builders building on AWS, Nova is worth comparing against Claude on Bedrock and open-weight alternatives like Llama 4. Nova models are tuned for AWS integration patterns and priced to be competitive within the Bedrock ecosystem. They support tool use, structured outputs, and the Bedrock Agent framework. The tradeoff versus Anthropic or OpenAI models is that Nova models are not as widely discussed in the developer community and third-party benchmark comparisons are less extensive.

This definition is AI-generated and refreshed weekly. It may contain inaccuracies. Use your own judgment, especially for production decisions.
Related terms
Inference APIRAGEmbeddingsVector database