From our custom-built Inferentia and Trainium chips, to offering best-in-class foundation models on Amazon Bedrock, and AI-powered experiences like Rufus and Alexa, we’re committed to delivering generative AI (Gen AI) solutions that offer real-world value to our customers. Our goal is to use AI to simplify the lives of shoppers, sellers, advertisers, enterprises, and everyone in between.
As the next step in our AI journey, we’ve built Amazon Nova, a new generation of foundation models (FMs). With the ability to process text, image, and video as prompts, customers can use Amazon Nova-powered generative AI applications to understand videos, charts, and documents, or generate videos and other multimedia content.
“Inside Amazon, we have about 1,000 Gen AI applications in motion, and we’ve had a bird’s-eye view of what application builders are still grappling with,” said Rohit Prasad, SVP of Amazon Artificial General Intelligence. “Our new Amazon Nova models are intended to help with these challenges for internal and external builders, and provide compelling intelligence and content generation while also delivering meaningful progress on latency, cost-effectiveness, customization, information grounding, and agentic capabilities.”
The new Amazon Nova models available in Amazon Bedrock include:
- Amazon Nova Micro, a text-only model that delivers the lowest latency responses at very low cost.
- Amazon Nova Lite, a very low-cost multimodal model that is lightning fast for processing image, video, and text inputs.
- Amazon Nova Pro, a highly capable multimodal model with the best combination of accuracy, speed, and cost for a wide range of tasks.
- Amazon Nova Premier, the most capable of Amazon’s multimodal models for complex reasoning tasks and for use as the best teacher for distilling custom models (available in the Q1 2025 timeframe).
- Amazon Nova Canvas, a state-of-the-art image generation model.
- Amazon Nova Reel, a state-of-the-art video generation model.
Amazon Nova Reel transforms a single image input into a brief video with the prompt: dolly forward.
How Amazon Nova models will benefit customers
All Amazon Nova models are incredibly capable, fast, cost-effective, and have been designed to be easy to use with a customer’s systems and data. They support a wide range of tasks across 200 languages and multiple modalities. Amazon Nova Micro, Amazon Nova Lite, and Amazon Nova Pro are at least 75 percent less expensive than the best performing models in their respective intelligence classes in Amazon Bedrock. They are also the fastest models in their respective intelligence classes in Amazon Bedrock.
The models are integrated with Amazon Bedrock, a fully managed service that makes high-performing FMs from leading AI companies and Amazon available for use through a single API. Using Amazon Bedrock, customers can easily experiment with and evaluate Amazon Nova models, as well as other FMs, to determine the best model for an application.
The models also support custom fine-tuning, which allows customers to point the models to examples in their own proprietary data that have been labelled to boost accuracy. The Amazon Nova model learns what matters most to the customer from their own data (including text, images, and videos), and then Amazon Bedrock trains a private fine-tuned model that will provide tailored responses.
In addition to supporting fine-tuning, the models also support distillation, which enables the transfer of specific knowledge from a larger, highly capable “teacher model” to a smaller, more efficient model that is highly accurate, but also faster and cheaper to run.
Amazon Nova models are integrated with Amazon Bedrock Knowledge Bases and excel at Retrieval Augmented Generation (RAG), which enables customers to ensure the best accuracy by grounding responses in an organisation’s own data.
Amazon Nova models have been optimised to make them easy to use and effective in agentic applications that require interacting with an organisation’s proprietary systems and data through multiple APIs to execute multistep tasks.
Creative content generation
With the output quality, intuitive API platform, and advanced customisation opportunities, the Amazon Nova creative generation models, Amazon Nova Canvas and Amazon Nova Reel, are helping Amazon Ads remove barriers for sellers and advertisers leading to new levels of creativity and innovation, like using video advertising for the first time, creating display and video campaigns for more products in their catalogue, and experimenting with new strategies like keyword-level creative. On average, we’ve seen brands using these tools advertise five times more products and twice as many images per advertised product, shifting budgets to what’s driving the best results.
In this example, Amazon Ads used Amazon Nova Reel to create a video advertisement for a fictional boxed pasta brand. In the whimsical “Pasta City,” buildings are sculpted from towering tubes of cannelloni noodles, the neighborhood is sprinkled with Italian spice landscaping, and the streets are lined with savory marinara sauce, fusilli noodles, and tender meatballs. The ad is an example of how advertisers can bring their products to life and create high-quality content using Amazon Nova models.
Video understanding
To put the video understanding capabilities of Amazon Nova Pro to the test, we asked the model to review and describe a brief silent video clip of a football game. The results include details about the setting of the game, the team uniforms, descriptions of actions taken by the players, and how the play culminates.
Video Prompt: “Describe the video.”
Output: “The video depicts a football game in progress on a green field. Players from two teams, one in yellow uniforms and the other in white, are engaged in a play. The quarterback from the yellow team throws a pass to a receiver, who catches the ball and begins running downfield. Defenders from the white team pursue him, attempting to tackle him. The play culminates in a tackle, bringing the receiver down on the field.”
Generating creative materials
Visual understanding can also help customers generate additional creative material. For example, Amazon Nova Pro can suggest several social media captions for the same video of the football game.
Looking ahead
We will introduce two additional Amazon Nova models in 2025, including a speech-to-speech model and a native multimodal-to-multimodal—or “any-to-any” modality model. Our speech-to-speech model will understand streaming speech input in natural language, interpreting verbal and nonverbal cues (like tone and cadence), and delivering natural humanlike interactions, while our any-to-any model will be capable of processing text, images, audio, and video, as both input and output. It will simplify the development of applications where the same model can be used to perform a wide variety of tasks, such as translating content from one modality to another, editing content, and powering AI agents that can understand and generate all modalities.
Responsible AI
Amazon Nova models are built with integrated safety measures and protections. The company has launched AWS AI Service Cards for Amazon Nova models, offering transparent information on use cases, limitations, and responsible AI practices.
This is only the beginning for Amazon Nova, and we’re excited to continue innovating to deliver real-world value to every Amazon customer. Learn more and get started with Amazon Nova.
