- AWS Nova Forge offers a new way to incorporate AI models
- Customers can pick from a range of Amazon models
- Nova 2 introduces a range of foundational models for usage too
AWS has revealed Nova Forge, a new service which will allow users to customize the company’s AI models exactly to their liking.
Announced at AWS re:Invent 2025, the platform will see customers build frontier models by first selecting a starter model and checkpoint, before mixing their data with Amazon-curated datasets, before deploying on Bedrock.
The company also revealed Amazon Nova 2 with four new foundation models, available in three tiers – scaling from everyday workloads up to more complex work, and a new speech-to-speech humanlike conversations.
Nova Forge
Available for $100,000 a year, Nova Forge will allow customers to create “Novellas” – custom versions of AWS Nova models.
These can be pre-trained, mid-trained or post-trained models, with companies then able to train on their own proprietary data, much cheaper than doing things from scratch, which can cost millions of dollars and take up thousands of workers hours.
“The more you customize models, the more you add a bunch of data in post training, these models tend to forget some of that interesting stuff that it learned earlier the core reasoning,” AWS CEO Matt Garman said in his keynote.
“It’s a little bit like humans trying to learn new language. When you start when you’re really young, it’s actually relatively easy to pick up, but when you try to, you learn a new language later in life, it’s actually much, much harder. Model training is kind of like this too.”
Nova was first announced in 2024, and Nova Forge is already being used across several internal Amazon teams.
Follow TechRadar on Google News and add us as a preferred source to get our expert news, reviews, and opinion in your feeds. Make sure to click the Follow button!
And of course you can also follow TechRadar on TikTok for news, reviews, unboxings in video form, and get regular updates from us on WhatsApp too.
