Amazon Nova Forge lets clients customize AI models for $100,000 a year
Amazon found a way to allow cloud clients to deeply customize generative AI models. The catch is that the system costs $100,000 annually.
Amazon Web Services’ Nova Forge offering gives organizations access to Amazon’s AI models at various stages of training, allowing them to incorporate their own data earlier into the process.
Companies can fine-tune large language models once they are already trained. Nova Forge’s results will rely more heavily on data provided by customers. Nova Forge customers will also have the option to develop vulnerability-weighted models, but this does not include training data and computing infrastructure.
Amazon said organizations that put together their own models could end up spending hundreds of millions or billions of dollars, meaning it’s more cost-effective to use Nova Forge.
AWS launched its own models under the Nova brand in 2024, but these are not the first choice of most software developers. One July questionnaire Amazon-backed Anthropic controls 32% of the enterprise LLM market, followed by OpenAI with 25%, Menlo Ventures said in the middle of this year. Google with 20% and Meta At 9%, Amazon Nova has less than a 5% share, a Menlo spokesman said.
Nova models, like Anthropic’s Claude 4.5 models, are available through AWS’s Bedrock service to run the models on Amazon cloud infrastructure.
“We’re a frontier lab focused on customers,” Rohit Prasad, Amazon’s chief artificial general intelligence scientist, told CNBC in an interview. “This is what our customers wanted, so we invented it for them to make it happen.”
Nova Forge is also used by internal Amazon customers, including teams working in the company’s stores and the Alexa AI assistant, Prasad said.
reddit An artificial intelligence model was needed to manage complex content about the many topics people discuss on the social network. Prasad said engineers found that the Nova model, developed with Reddit data through Forge, outperformed commercially available large-scale models. Amazon, Booking.com, Nimbus Therapeutics, Nomura Research Institute and Sony have also built models with Forge, he said.
Organizations can ask Amazon engineers to help them build Forge models, but that help isn’t included in the new service’s $100,000 annual fee.
AWS is also introducing new models for developers at its Reinvent conference in Las Vegas this week.
Amazon said there’s a reasoning model that tests of the Nova 2 Pro show it performs at least as well as Anthropic’s Claude Sonnet 4.5, OpenAI’s GPT-5 and GPT-5.1, and Google’s Gemini 3.0 Pro Preview. Reasoning involves executing a series of calculations that may take extra time in response to requests to produce better answers. Prasad said that Nova 2 Pro will be available in early access to AWS customers with Forge subscriptions. This means Forge customers and Amazon engineers can try out Nova 2 Pro at the same time.
Nova 2 Omni is another reasoning model that can process incoming images, speech, texts and videos and produce images and text. Amazon said this is the first reasoning model with this range of capabilities. Amazon hopes that by offering a versatile model, it can reduce the cost and complexity of incorporating AI models into applications.
Tens of thousands of organizations use Nova models every week, Prasad said. AWS said it has millions of customers. Nova is the second most popular model family at Bedrock, Prasad said. The top model group is from Anthropic.
WRISTWATCH: Amazon will kick off its AI conference next week: Maxim’s Tom Forte on what to expect
