Save StorySave this storySave StorySave this story
Amazon has announced a new family of frontier artificial intelligence models—and a new way for customers to build frontier models of their own.
The ecommerce giant announced the second generation of its Nova AI models at re:Invent, a company conference held in Las Vegas. The models are nowhere near as popular as those offered by rivals like OpenAI and Google, but Amazon’s plan to make them highly customizable could see them gain traction with its cloud users.
Amazon detailed two improved large language models, Nova Lite and Nova Pro, a new real-time voice model called Nova Sonic, and a more experimental model called Nova Omni that performs a simulated kind of reasoning using images, audio, and video as well as text. The new models are being made available today to a limited number of customers.
More significantly, given the importance of its cloud business, Amazon is releasing a tool called Nova Forge that will let customers create specialized frontier models by adding their own training data to unfinished versions of the Nova 2 Lite and Pro models.
It is already possible to fine-tune off-the-shelf AI models like Google’s Gemini and OpenAI’s GPT. But Amazon’s approach lets customers add data at various stages of model training, including the process of building the base model, a stage known as custom pretraining that is normally reserved for large AI labs.
“Everyone is looking for a frontier model that's an expert in their domain,” Rohit Prasad, who leads Amazon’s AI efforts, told WIRED ahead of today’s announcements. Prasad says that Amazon developed the technologies behind Nova Forge to empower internal teams, including those developing Alexa and AI agents to build custom models. “This is essentially a new open training paradigm,” he says.
One customer that has tested the approach is Reddit, which used Nova Forge to create a custom model to identify content that breaks the platform’s rules.
Fine-tuning a conventional model would not work, says Reddit chief technology officer Chris Slowe, because most models are designed to avoid offensive or violent content entirely, meaning they would refuse to analyze some materials. Slowe says that custom pre-training, combined with conventional fine-tuning, produced a frontier model that is expert at understanding and using Reddit.
“Other LLMs understand Reddit as a concept, and how Reddit works, but they're not down in the weeds,” Slowe says. “We really built a Reddit expert model.”
Slowe adds that Reddit’s customized model could have a range of uses, and will most likely be put to work next to automate content moderation.
Other companies testing Nova Forge include Booking.com, Sony, and Nimbus Therapeutics, a biotech firm.
Allowing customers to craft specialized models may prove savvy as companies look for tools that go beyond the capabilities of the latest general-purpose models. About three-quarters of US companies see AI as a high priority, according to a survey from the consulting company Bain released in November. Those companies also, however, report a wide range of problems in using AI, among them a lack of expertise and resources needed to build custom models.
Today most AI models are either closed, meaning they can only be accessed through API or app, or open, meaning they can be downloaded and run on one’s own hardware. Many companies opt to work with open models—the most popular being ones that come from Chinese companies, like Alibaba and DeepSeek—because they are cheaper to experiment with and can be modified with relative ease. The data used to train these open models is typically not released, however, which limits and complicates the process of tuning them.
Nova Forge offers a new approach, albeit one that is locked into Amazon’s cloud. Building a large language model from scratch can cost tens or hundreds of millions of dollars. Prasad says a frontier model built using Nova Forge should be significantly cheaper, without providing specifics.
Amazon remains something of a dark horse in the AI race, having been relatively late to develop truly cutting-edge AI language models. The company is, however, quietly building up a portfolio of advanced AI capabilities. It has also integrated generative AI into its shopping platform, for example through an ecommerce-focused chatbot helper called Rufus.
Like other big tech companies, Amazon is investing billions to build new AI infrastructure, part of a colossal—and potentially risky—bet that demand for AI will continue to grow at an impressive clip.
Amazon is competing with Google and Microsoft for cloud customers. OpenAI is also rapidly building its own infrastructure and could someday become a cloud player itself. It has hedged its bets by investing $8 billion in Anthropic, a major competitor to OpenAI founded by staff who left the maker of ChatGPT. Amazon is also looking to challenge Nvidia’s hardware dominance; Anthropic’s latest models are trained on Amazon’s custom Trainium chips.
Amazon says Nova 2 Pro matches or exceeds OpenAI’s GPT-5 and GPT-5.1, Google’s Gemini Pro 2.5 and Gemini 3.0 Pro, and Sonnet 4.5 from Anthropic across a range of benchmarks. Prasad notes that the model is especially good at agentic tasks like following complex instructions and using tools on a computer. The company says that its smaller model, Nova 2 Lite, is similar to Claude 4.5 Haiku, GPT-5 Mini, and Gemini Flash 2.5 on various benchmarks.
Nova 2 Omni shows that Amazon is no slouch in AI research these days. A fully multimodal reasoning model, it can take images, audio, and video as well as text as input and can perform simulated reasoning to generate output. Prasad says to his knowledge no other AI company has released a fully mutli-model of this kind.
Reddit’s Slowe says the customizable nature of Nova is probably its most important quality. “I do believe it has a lot of potential,” he says. “For a large set of situations, it will be substantially better than what we get off the shelf.”
