Amazon Web Services is rolling out a series of new homegrown AI models and services for enterprise customers to build their own custom versions. The cloud provider announced Nova 2, which consists of four new AI models in the Nova model family, during AWS CEO Matt Garman’s AWS re:Invent keynote on Tuesday.
The first version of AWS Nova was announced last year at the company’s annual technology conference. At that time, the company released four text generation models and one image generation model. This year, AWS is upgrading its model and launching accompanying services.
“The momentum is really impressive,” Garman said in his keynote address Tuesday. “Nova has been and continues to be used by tens of thousands of customers and is growing. Everyone from big marketing companies to technology leaders like Infosys, Blue Origin and Robinhood to innovative startups like NinjaTech AI. And now we’re making Nova even better.”
The four new models include Nova 2 Lite, a more cost-effective inference model. Before responding, reasoning AI models can process “thoughts”, text, images, and videos to generate text suitable for everyday tasks. Nova 2 Pro is a reasoning agent that can process text, images, video, and audio, designed for “highly complex tasks” such as coding.
Nova 2 Sonic is a new speech synthesis model used for conversational AI. Nova 2 Omni is a multimodal inference and generation model that can process image, text, video, and audio input and generate both text and images.
Alongside the model upgrade, AWS also announced a new service called Nova Forge that allows AWS cloud customers to build their own frontier versions of AWS Nova models called Novellas for $100,000 per year, as reported by CNBC. The service allows companies to access pre-, during-, or post-training models and train them on their own data.
Garman said this could solve some of the problems that arise when companies try to incorporate their data into already trained AI models.
tech crunch event
san francisco
|
October 13-15, 2026
“The more you customize your models, the more you add a lot of data after training, the more these models tend to forget some of the interesting things they learned earlier in their core inference,” Garman said. “It’s a bit like humans trying to learn a new language. If you start young, it’s actually relatively easy to learn, but if you try to learn a new language later on, it’s actually much harder. Training a model is similar.”
Companies like Reddit, Sony, and Booking.com are early customers of Nova Forge.
Check out all of our coverage of TechCrunch’s annual enterprise technology event here.
Catch the latest announcements on everything from agent AI to cloud infrastructure to security and more from Amazon Web Services’ flagship event in Las Vegas. This video is provided in partnership with AWS.
