Hajra Naz
Amazon Web Services is expanding its AI offerings. The company rolled out a new set of homegrown AI models. It also launched a service that lets enterprise customers build their own custom versions. This announcement came during AWS CEO Matt Garman’s AWS re:Invent keynote on Tuesday.
The highlight is Nova 2, a fleet of four new AI models added to the Nova model family. AWS first introduced Nova last year at re:Invent. The initial release included four text-generating models and one image-generating model. This year, AWS upgraded the models and added an accompanying enterprise service. Trending keywords for this topic include AWS Nova AI, enterprise AI customization, multimodal AI, and AI model training.
AWS Nova Growth and Adoption
“The momentum has been really fantastic,” Garman said during his keynote. Nova has grown to be used by tens of thousands of customers today. This includes marketing giants, tech leaders like Infosys, Blue Origin, and Robinhood, and innovative startups like NinjaTech AI. Today, we’re making Nova even better.”
Nova AI is now widely adopted across industries. It powers marketing campaigns, enterprise applications, and AI research projects. Its adoption signals AWS’s growing influence in the enterprise AI market.
Read More: Amazon Launches Smart AI Assistant for Marketplace Seller Support
The Four New Nova 2 Models
AWS introduced four new models with distinct capabilities:
-
Nova 2 Lite – A cost-effective reasoning model. It “thinks” before responding. It can process text, images, and video to generate useful outputs for everyday business tasks.
-
Nova 2 Pro – A reasoning agent for highly complex tasks. It handles text, images, video, and speech. Ideal for coding, research, and technical workflows.
-
Nova 2 Sonic – A speech-to-speech model designed for conversational AI applications. It enables real-time voice interaction and dialogue generation.
-
Nova 2 Omni – A multimodal reasoning and generation model. It can process text, images, video, and speech input. It produces both text and image outputs.
These models strengthen AWS’s position in multimodal AI, conversational AI, and enterprise AI automation.
Nova Forge: Custom AI for Enterprises
Alongside the model upgrades, AWS announced Nova Forge, a new service for enterprises. It allows companies to build their own frontier version of Nova models, called Novellas. The service costs $100,000 per year, according to CNBC.
Nova Forge supports multiple stages of AI model access:
-
Pre-trained models
-
Mid-trained models
-
Post-trained models
Companies can then train these models on their proprietary data. This allows for enterprise-grade AI customization, preserving critical reasoning while incorporating company-specific datasets.
The Challenge of Model Customization
Garman explained why this is important. “The more you customize models, the more you add data in post-training. Sometimes the model forgets earlier reasoning. It’s like humans learning a new language later in life. Models can struggle to retain earlier knowledge if not carefully managed.”
This highlights a key trend in AI development: balancing pre-trained knowledge with enterprise-specific customization. Nova Forge aims to solve this challenge, making AI more adaptable for business use.
Read More: AWS to Launch AI Agent Marketplace Next Week with Anthropic Partnership
Early Adopters
Several high-profile companies are already using Nova Forge. Early customers include Reddit, Sony, and Booking.com. These companies aim to integrate custom AI models into operations, customer engagement, and internal workflows.
AWS’s Nova 2 models, along with Nova Forge, reflect a growing enterprise demand for customizable, multimodal AI. They also show how cloud providers are offering tools to make AI more controllable, scalable, and secure for enterprise applications.
