AWS Unveils Nova 2 AI Model Family with Multimodal Omni & Agentic Nova Act

AWS announced the debut of its Nova 2 AI model family at re:Invent 2025, unveiling four variants—Lite, Pro, Sonic and Omni—alongside Nova Forge, an open-training service, and Nova Act, a platform for creating autonomous web‑browsing AI agents. This update enables enterprises to customise and deploy AI systems that mirror their own operational logic, with the ecosystem already being adopted by Cisco, Siemens, Reddit, Sony and others.

Nova 2 Lite is described as a cost‑efficient reasoning model capable of handling text, images and video, and, according to AWS, it surpasses Claude Haiku 4.5 and GPT-5 Mini in both performance and price efficiency; Nova 2 Pro is positioned as Amazon’s most powerful in‑house reasoning model to date, while Nova 2 Sonic focuses on speech‑to‑speech interactions and multilingual conversation.

Nova 2 Omni is billed as the industry’s first single model capable of fully multimodal processing and generation, able to analyse up to 750,000 words, hours of audio, or extended video content. Reddit CTO Chris Slowe noted that Nova Forge allows the company to unify content‑moderation systems, and Nova Act is built atop a customised Nova 2 Lite model to perform browser‑level actions, achieving 90% reliability in early workflow trials. 4 December 2025.