Amazon has announced a series of new AI foundation models under a new “Nova” branding that will be available as part of the Amazon Bedrock model library in AWS.
The company is also training a model called Amazon Nova Premier, which it says will be “our most capable multimodal model for complex reasoning tasks.” Amazon aims to make Nova Premier available in “early 2025.”
Amazon is releasing content generation models, too: Amazon Nova Canvas, an image generation model, and Amazon Nova Reel, a video generation model. The company says that these models have “watermarking capabilities” to “promote responsible AI use.” As an example of what’s possible with Nova Reel, Amazon has shared this mock ad for a fake pasta brand.
Later in 2025, Amazon plans to release a speech-to-speech model and “a native multimodal-to-multimodal” model, according to a blog post.
Amazon announced these new models at its AWS re:Invent conference, which is happening now in Las Vegas. At the show, the company also said that it’s building a huge AI compute cluster that relies on its Trainium 2 chips in partnership with Anthropic (which it has invested $8 billion in). “When completed, it is expected to be the world’s largest AI compute cluster reported to date available for Anthropic to build and deploy its future models on,” according to Amazon.
The company, like many other big tech players, is racing to release new AI products and features to stay ahead of newer companies like OpenAI. Where Amazon could have an advantage is how much internet infrastructure is already powered by AWS — large enterprises may be more willing to use Amazon’s AI offerings because the company has already a trusted reputation. An Apple exec even appeared onstage today at re:Invent to talk about how the company relies on Amazon’s custom AI chips.
