Today, during the AWS Re:Invent 2025, Amazon announced its new Nova 2 models. Promising to provide industry-leading price-performance across reasoning, multimodal processing and conversational AI, four Nova 2 models have been announced.
As part of the new family of Nova 2 models, Amazon unveiled Lite, Pro, Sonic and the brand new Omni model for this series. Each one is calibrated for varying uses of AI. Amazon is also launching a new “open training” service so organizations can build their own custom models on Nova.
Starting with Nova 2 Lite, Amazon reveals that this is the new, most cost-effective reasoning model for everyday workloads. Nova 2 Lite is designed to process text, images, and videos to generate text. It’s built for “industry-leading price performance,” superseeding 13 out of 15 benchmarks when compared to Claude Haiku 4.5. Nova 2 Lite has capabilities across documents, videos and can generate code.
The next step up is Nova 2 Pro, which is “Amazon’s most intelligent reasoning model.” Much like the Lite model, Pro can process text, images, and videos, but can also process speech to text. As a more advanced model, Nova 2 Pro is built for complex tasks like agentic coding, long-range planning and problem solving on a higher level. When compared to GPT-5.1, Nova 2 is equal to or better on 8 out of 16 benchmarks. On top of standardized capabilities, Nova 2 Pro is designed for video reasoning, solving advanced math equations, and executing agentic and software engineering tasks.
Both Nova 2 Lite and Nova 2 Pro have built-in web browsing so they can search the web or current information for the user. They can also run code directly through the models, so both can stay up to date on information served to the end user.
Nova 2 Sonic is Amazon’s new speech-to-speech model, unifying text and speech in real time for human-like conversations. Nova 2 Sonic is said to offer multilingual support with expressive voices. It also offers a one-million token context window for sustained interactions. Amazon has also built it with the capabilities of switching between voice and text. This model can integrate with Amazon Connect, telephony providers like AudioCodes Codes and conversational AI like LiveKit or Pipecat.
Finally, and most interestingly, Nova 2 Omni has been announced. This is a unified multimodal reasoning and generation model, capable of processing text, images, video, and speech inputs while also generating text and images. This is an industry first. The model is capable of handling 750,000 words, hours of audio, long videos, and hundred-page documents while analyzing entire product catalogues and video libraries.
Amazon confirms that organizations like Cisco, Siemens, Sumo Logic, and Trellix are already using Nove 2 models, running them on applications from everything from agentic threat detection to video understanding.
Last up, Amazon reveals Nova Forge, a way for organizations to build their own optimized variants of Nova. These are being called ‘Novellas’ and blend proprietary data with Nova’s capabilities. This service is being referred to as open training and gives organizations access to pre-trained, mid-trained, and post-trained Nova model checkpoints. Customers can start building their own Novellas with Nova 2 Lite today. Nova Forge customers will get access to Nova 2 Pro and Nova 2 Omni.
Partnering organizations include Booking.com, Cosine AI, Nimbus Therapeutics, Nomura Research Institute, OpenBabylon, Reddit, and Sony. All of these are building their own models with Nova Forge.
Want to see more of our stories on Google?
P.S. Want to keep this site truly independent? Support us by buying us a beer, treating us to a coffee, or shopping through Amazon here. Links in this post are affiliate links, so we earn a tiny commission at no charge to you. Thanks for supporting independent Canadian media!
