If you've ever tried to train a model with your domain specific data, you know how difficult it is to push the model to behave in ways that diverge from their core training data. Nova Forge intends to change that.
A first-of-its-kind service, Nova Forge lets enterprises build custom frontier models that incorporate both Nova’s strengths and their own proprietary expertise. These tailored variants—called “Novellas”—are optimized versions of Nova that can be designed and built around each organization’s unique needs.
Novellas marry Nova’s expansive knowledge and reasoning with tailored, business-specific understanding. Novellas will give companies a model that truly reflects how they operate.
Why do enterprises need Nova Forge?
Generic foundation models typically struggle with the precision and domain specificity required in enterprise environments. Even when companies fine-tune open-source models, accuracy can degrade as more data is introduced, leading to brittle or inconsistent behavior.
Enterprises embedding proprietary intelligence into AI systems face three major constraints -
- Surface-level fine-tuning of commercial models, which fails to deeply encode domain expertise.
- Continuing to train open-weights models without access to their foundational data, risking erosion of baseline abilities like reasoning and instruction adherence.
- Developing LLMs from scratch, an approach that demands massive infrastructure, capital, and specialized talent.
Nova Forge Features
It introduces a new approach called “open training,” which provides customers with exclusive access to pre-trained, mid-trained, and post-trained Nova checkpoints. This allows enterprises to combine their proprietary data with Amazon’s curated Nova datasets at every stage of the training process.
Beyond model checkpoints and data-mixing capabilities, Nova Forge also offers three additional capabilities -
- Ability to train AI using your own environments – Gyms - simulated scenarios that reflect their real-world use case.
- Synthetic data-based distillation - to create smaller, faster models that maintain their intelligence at a lower cost.
- Access to a responsible AI toolkit - to implement safety controls.
First Impressions of Nova Forge
The new service costs $100,000 per year and doesn’t include help from Amazon experts. While this is still way cheaper than the cost of building a LLM from scratch (hundreds of millions of dollars), this pricing range might attract only a niche target audience.
Amazon engineers can assist with building Forge models upon request, though this support falls outside the service’s $100,000 yearly subscription.
Nova Forge Use Cases
- Organizations like Booking.com, Cosine AI, Nimbus Therapeutics, Nomura Research Institute, OpenBabylon, Reddit, and Sony are building their own models with Nova Forge to better serve their unique requirements.
- “Working with Nova Forge is allowing us to improve content moderation on Reddit with a more unified system that's already delivering impressive results,” said Chris Slowe, CTO, Reddit.
- Nova Forge is also in use by internal Amazon customers, including teams that work on the company’s stores and the Alexa AI assistant, Rohit Prasad, Amazon Head Scientist of AI had said.
- Sony said it was putting Nova Forge in the middle of its agentic AI architecture, which is built on AgentCore.
What's next for Nova Forge?
Enterprises can already use Nova 2 Lite to construct their first Novellas. Through Nova Forge, they also gain early access to the more powerful Nova 2 Pro and Nova 2 Omni models—accelerating their ability to prototype, refine, and deploy higher-capability Novellas and next-generation applications.
This article is part of our comprehensive guide to AWS AI. Explore the guide to know more about AWS AI tools like Bedrock and SageMaker, AWS AI LLMs like Nova 2, Nova 2 Omni and why AWS AI infrastructure is the best way to build gen AI based solutions that can scale in production.
