Amazon has launched a comprehensive suite of artificial intelligence models called Nova, marking a significant step in the company's generative AI strategy at its AWS re:Invent conference in Las Vegas.

The Nova family includes three primary understanding models currently available: Nova Micro, a text-only model optimised for speed and cost; Nova Lite, a low-cost multimodal model processing images, video, and text; and Nova Pro, a highly capable multimodal model. A more advanced model, Nova Premier, is slated for release in early 2025.

Rohit Prasad, senior vice president of Amazon Artificial General Intelligence, explained the motivation behind the new models: "Inside Amazon, we have about 1,000 generative AI applications in motion, and we've had a bird's-eye view of what application builders are still grappling with."

Alongside understanding models, Amazon has introduced content generation models Nova Canvas for images and Nova Reel for video. The company emphasised built-in safety features, with watermarking capabilities to promote responsible AI use.

Amazon plans to expand its AI capabilities further, with a speech-to-speech model and a "multimodal-to-multimodal" model planned for 2025. These models aim to process and generate content across text, images, audio, and video.

The announcement is part of Amazon's broader AI strategy, which includes building a massive AI supercomputer in collaboration with Anthropic. The company has invested $8 billion in Anthropic and is constructing what it claims will be the world's largest AI compute cluster using its custom Trainium 2 chips.

Matt Garman, chief executive officer of Amazon Web Services, highlighted the company's focus on practical AI solutions. "Many customers are far less interested in pushing the frontier of generative AI than in finding ways to make the technology cheaper and more reliable," he told Wired.

The Nova models are integrated with Amazon Bedrock, allowing customers to experiment with and fine-tune AI models using their proprietary data. Amazon claims the models are at least 75 per cent less expensive than comparable models in their intelligence classes.

The company is positioning itself as a serious competitor in the AI landscape, leveraging its extensive cloud infrastructure and custom silicon to offer unique AI solutions to businesses and developers.

---