Amazon described these models as tools to enhance productivity and creativity across industries, offering advanced multimodal capabilities, cost-efficiency, and speed. Nova models are available through Amazon Bedrock, a fully managed AI service that simplifies access to high-performing AI tools.
The Nova lineup includes a variety of models tailored to specific use cases:
- Nova Micro delivers low-latency, text-only responses at minimal cost.
- Nova Lite processes text, images, and videos quickly and affordably.
- Nova Pro balances accuracy, speed, and cost for diverse applications.
- Nova Premier, expected in 2025, will cater to complex reasoning tasks.
- Nova Canvas and Nova Reel specialize in generating high-quality images and videos, respectively.
These models are designed to integrate easily with businesses’ proprietary systems and data. They support fine-tuning for improved accuracy, allowing organizations to customize outputs based on their unique requirements. Features like knowledge grounding through retrieval-augmented generation (RAG) and support for multistep tasks make the Nova models suitable for a wide range of applications, from data analysis to creative content generation.
Additionally, Amazon Nova models are optimized for efficient agentic applications, enabling them to interact with multiple APIs and execute complex workflows. For creative industries, tools like Nova Canvas and Nova Reel offer opportunities to streamline the production of high-quality visuals and videos, unlocking new possibilities for advertising and storytelling.
Looking ahead, Amazon plans to expand the Nova family with two new models in 2025. These will include a speech-to-speech model for natural, humanlike interactions and an “any-to-any” model capable of processing multiple modalities—text, images, audio, and video—both as inputs and outputs.
Amazon has emphasized responsible AI practices, integrating safety measures and releasing AI Service Cards to provide transparency about the models’ capabilities and limitations.
