Amazon has new models for artificial intelligence — and a way for customers to build their own

Amazon She announced a new family of borders artificial intelligence Models – A new way for customers to build their own parametric models.

The e-commerce giant announced the second generation of Nova AI models at the re:Invent conference held in Las Vegas. The models aren’t as popular as those offered by competitors like OpenAI and Google, but Amazon’s plan to make them highly customizable could see them gain traction with cloud users.

Amazon has detailed two models optimized for larger languages, the Nova Lite and Nova Pro; A new real-time audio model called NovaSonic; And a more experimental model called Nova Omni which simulates a type of reasoning using images, audio and video in addition to text. Today the new models are made available to a limited number of customers.

More importantly, given the importance of its cloud business, Amazon is also launching a tool called Nova Forge that will allow customers to create specialized parametric models by adding their own training data to unfinished versions of the Nova 2 Lite and Pro models.

It is already possible to improve off-the-shelf AI models such as Google’s Gemini and OpenAI’s GPT. But Amazon’s approach allows customers to add data at different stages of model training, including the process of building the basic model, a stage known as custom pre-training that is typically limited to large AI labs.

“Everyone is looking for a leading model who is an expert in their field,” Rohit Prasad, who leads Amazon’s AI efforts, told WIRED ahead of today’s announcements. Amazon developed the technologies behind Nova Forge to enable internal teams, including those developing Alexa and AI agents, to build custom models, Prasad says. “This is basically a new open training model,” he says.

One client who has already tested this approach is Redditwhich used Nova Forge to create a custom form to identify content that violates the platform’s rules.

Traditional model policing won’t work, says Chris Slough, Reddit’s chief technology officer, because most models are designed to avoid offensive or violent content altogether, which means they’ll refuse to analyze some material. Custom pre-training, combined with traditional fine-tuning, produced a leading model that is expert at understanding and using Reddit, Slow says.

“Other LLM students understand Reddit as a concept, and how Reddit works, but they’re not stuck,” Slough says. “We’ve already built an expert model on Reddit.”