Amazon Web Services (AWS), the cloud computing division of the tech giant, introduced the Nova family of artificial intelligence (AI) models on Tuesday at its ongoing re:Invent conference. There are five different large language models (LLMs) under the Nova branding, with three of them capable of only text generation. Apart from this Nova also includes an image-generation model and a video-generation model. The company stated that the new generation of AI models comes with improved intelligence and competitive pricing, and is currently available on Amazon Bedrock.
AWS Introduces Nova AI Models
In a post, Amazon detailed the new generation of AI models. Currently, five different LLMs have been introduced as part of the Nova series, and company CEO Andy Jassy highlighted that a sixth AI model dubbed Nova Premier will be launched in 2025.
Among the five models, three — Nova Micro, Nova Lite, and Nova Pro — can only generate text. However, there are differences between the three models. Micro only accepts text as input and provides the lowest latency responses in the entire series. It has a context window of 1,28,000 tokens.
On the other hand, the Nova Lite accepts images, videos, and text as inputs but only generates text. Nova Pro is the most capable multimodal AI model in the trio and can complete a wider range of tasks compared to the other two. Both of these models have a context window of 3,00,000 tokens.
Apart from these, there are two more models in the Nova series that Amazon calls “creative content generation models”. First is Nova Canvas, an image generation model that accepts text and images as inputs. The company has touted it as a tool for advertising, marketing, and entertainment.
Finally, the Nova Reel is a video generation model that can generate short videos from text and image prompts. It also gives users the option to control camera motion with natural language prompts. All of these models are available for the company’s enterprise clients and can be availed from the Amazon Bedrock platform.