AWS Unveils Nova 2 AI Models at re:Invent

Amazon Web Services unveiled a series of advancements in artificial intelligence during its annual re:Invent conference in Las Vegas this week, expanding its Nova foundation model platform with new reasoning and multimodal capabilities designed to help enterprises build and deploy custom AI solutions.

The centerpiece of the announcements is the Nova 2 family of models, which includes four variants tailored for different workloads. Nova 2 Lite serves as a cost-effective option for everyday tasks, processing text, images and videos to generate responses, while outperforming competitors on most benchmarks for document analysis, video understanding and code generation. Nova 2 Pro offers advanced reasoning for complex activities such as coding, long-range planning and problem-solving, handling text, images, videos and speech inputs with built-in web browsing and code execution features. Nova 2 Sonic introduces a speech-to-speech model that enables real-time, human-like conversations with multilingual support, expressive voices and a one-million-token context window, allowing seamless switches between voice and text. Completing the lineup is Nova 2 Omni, a multimodal model that...