At the Amazon Cloud Technology 2025 re:Invent global conference, Amazon announced a comprehensive expansion of its Nova product portfolio - launching four new Nova 2 models and multiple services. It announced that the Nova 2 model achieves industry-leading price-performance ratio in reasoning, multi-modal processing, conversational AI, code generation and Agent tasks.
Among them, Nova 2 Lite is a fast and economical inference model for daily workloads that can process text, image and video input and generate text output. The model was tied or better in 13 of 15 benchmarks compared to the Claude Haiku 4.5, tied or better in 11 of 17 benchmarks compared to the GPT-5 Mini, and tied or better in 14 of 18 benchmarks compared to the Gemini Flash 2.5. Nova 2 Lite is particularly outstanding in its ability to process various types of documents, extract key information from videos, generate code, provide accurate fact-based answers, and automate multi-step agent workflows.
Nova 2 Pro is an intelligent inference model that can process text, image, video and voice input and generate text output. It is ideally suited for highly complex tasks that require the highest accuracy, such as agent programming, long-term planning, and complex problem solving. In public benchmark tests, Nova 2 Pro was on par or better than Claude Sonnet 4.5 in 10 of 16 reviews; it was on par or better than GPT-5.1 in 8 of 16 reviews; and it was on par or better than Gemini 3 Pro Preview in 8 of 18 reviews.
Both Nova 2 Lite and Nova 2 Pro have built-in web page search and code execution capabilities. They can search the latest Internet information and run code directly as needed, ensuring that the model's answers are always based on the latest facts, rather than relying solely on training data. In addition, Amazon also launched new models such as Nova 2 Sonic and Nova 2 Omni.
Currently, tens of thousands of companies are already using Nova to support a variety of application scenarios, such as producing high-quality content, automating multi-step tasks, and accelerating the development of AI Agents.