rewrite this title Amazon Expands Nova AI Portfolio With Four AI Models And Advanced Tools For Custom AI Development

by
Alisa Davidson

Published: December 03, 2025 at 5:54 am Updated: December 03, 2025 at 5:54 am

by Ana

Edited and fact-checked:
December 03, 2025 at 5:54 am

To improve your local-language experience, sometimes we employ an auto-translation plugin. Please note auto-translation may not be accurate, so read original article for precise information.

Multinational technology company Amazon announced the expansion of its Nova portfolio with four new models, an “open training” service enabling organizations to create custom variants of Nova, and a service for building highly reliable AI agents. Nova is already used by tens of thousands of companies for applications such as generating high-quality content, automating complex workflows, and accelerating AI agent development. The new Nova 2 family aims to balance speed, cost, and intelligence across multiple AI domains.

Nova 2 Lite is a fast, cost-efficient reasoning model capable of processing text, images, and video to generate text. Users can adjust the model’s step-by-step reasoning to balance depth, speed, and cost, making it suitable for customer service chatbots, document processing, and business automation. Nova 2 Lite demonstrates strong performance across benchmarks compared to models such as Claude Haiku 4.5, GPT-5 Mini, and Gemini Flash 2.5, excelling in document processing, video information extraction, code generation, grounded question answering, and multi-step workflow automation.

Nova 2 Pro is Amazon’s most capable reasoning model, processing text, images, video, and speech for highly complex tasks such as agentic coding, long-term planning, and advanced problem-solving. It can also act as a “teacher” for knowledge distillation, transferring expertise to smaller, task-specific models. Nova 2 Pro shows strong performance compared to Claude Sonnet 4.5, GPT-5.1, Gemini 2.5 Pro, and Gemini 3 Pro Preview, particularly in multi-document analysis, video reasoning, complex instruction following, advanced mathematics, and software engineering tasks. Both Nova 2 Lite and Nova 2 Pro include web grounding and code execution, enabling them to provide up-to-date, fact-based responses and execute code directly.

Nova 2 Sonic is a speech-to-speech model that integrates text and speech understanding for real-time conversational AI, supporting multilingual, expressive voices and a one-million token context window. It allows seamless switching between voice and text and handles asynchronous tasks, enabling continuous conversation while actions complete in the background. Nova 2 Sonic integrates with Amazon Connect, telephony providers, and conversational AI frameworks, making it suitable for customer service, AI assistants, and interactive voice experiences, offering competitive performance against OpenAI’s gpt-realtime and Gemini 2.5 Flash models.

Nova 2 Omni is a unified multimodal model capable of processing text, images, video, and speech while generating both text and images. It can analyze large-scale documents, audio, and video simultaneously, streamlining workflows such as marketing campaigns by generating headlines, copy, social posts, and visuals in a single process. Nova 2 Omni demonstrates leading performance in multimodal reasoning benchmarks and high-quality image generation.

Organizations including Cisco, Siemens, Sumo Logic, and Trellix are using Nova 2 models for applications ranging from agentic threat detection to video understanding and voice AI assistants.

Nova Forge And Nova Act To Enable Custom AI Models And Browser-Based Agent Automation

Amazon has introduced Nova Forge and Nova Act to provide organizations with advanced tools for creating and deploying AI solutions. Organizations seeking to integrate proprietary knowledge into AI applications often face trade-offs, such as limited customization of existing models, training open-weight models without access to original data, or building models from scratch at high cost. Nova Forge addresses these challenges by enabling organizations to develop their own optimized variants of Nova, known as “Novellas,” combining proprietary data with Amazon’s advanced model capabilities. The service offers access to pre-trained, mid-trained, and post-trained Nova model checkpoints, allowing data integration at every stage of training. Customers can begin creating Novellas with Nova 2 Lite and receive early access to Nova 2 Pro and Nova 2 Omni for more advanced applications.

In addition to data-mixing and model checkpoints, Nova Forge allows training in custom reinforcement learning environments, creating smaller, efficient models through synthetic data-based distillation, and using a responsible AI toolkit for safety controls. Organizations such as Booking.com, Cosine AI, Nimbus Therapeutics, Nomura Research Institute, OpenBabylon, Reddit, and Sony are already leveraging Nova Forge to develop models tailored to their specific needs. Once a model is created, it can be deployed on Amazon Bedrock, providing enterprise-grade security, scalability, and data privacy while maintaining exclusive control over the customized model.

Nova Act is available on AWS for building and deploying AI agents that can perform tasks in web browsers. Powered by a custom Nova 2 Lite model, Nova Act allows rapid creation and management of fleets of agents that automate browser-based workflows, achieving early customer reliability of 90% and outperforming competing models. The system trains agents through reinforcement learning in simulated web environments, enabling high performance on tasks such as CRM updates, website testing, and form submissions.

Developers can prototype agents in a no-code environment using natural language prompts, refine them in familiar IDEs, and deploy them on AWS with full management and monitoring via the Nova Act console. Organizations across industries have already seen significant results: Sola Systems automates hundreds of thousands of workflows per month, 1Password simplifies user logins across multiple websites, Hertz accelerates software delivery and QA processes fivefold, and Amazon Leo reduces QA testing from weeks to minutes while increasing execution speed. These solutions demonstrate how Nova Forge and Nova Act allow organizations to efficiently build, customize, and scale AI applications with enterprise-grade security and operational control.

Alisa, a dedicated journalist at the MPost, specializes in cryptocurrency, zero-knowledge proofs, investments, and the expansive realm of Web3. With a keen eye for emerging trends and technologies, she delivers comprehensive coverage to inform and engage readers in the ever-evolving landscape of digital finance.

More articles