AWS announces Nova 2 AI models to rival Google, OpenAI
What's the story
Amazon Web Services (AWS) has announced a new range of artificial intelligence (AI) models and a service aimed at giving enterprise customers more control. The new models are part of the Nova family, which was first introduced last year with four text-generating and one image-generating model. The latest addition to this lineup is Nova 2, comprising four new AI models.
Model breakdown
Nova 2 models: A closer look
The Nova 2 family includes Micro, Lite, Pro, and Premier (coming soon) variants. Nova 2 Lite, a cost-effective reasoning model that can handle text, images, and videos to generate everyday task-oriented responses. Then there's Nova 2 Pro, a more advanced reasoning agent capable of handling highly complex tasks like coding. The new lineup also features Nova 2 Sonic for conversational AI and Nova 2 Omni, a multimodal reasoning and generation model that can process multiple types of inputs.
Customization service
Nova Forge: A new service for enterprises
Along with the model upgrades, AWS also unveiled Nova Forge. This new service lets AWS cloud customers create their own versions of the Nova models, called Novellas, for an annual fee of $100,000. The offering gives enterprises access to pre-trained, mid-trained or post-trained models that they can then train on their own proprietary data.
Problem-solving approach
Nova Forge addresses data integration challenges
Garman emphasized that Nova Forge could solve problems enterprises face when trying to integrate their own data into already-trained AI models. He explained this by comparing it to humans learning a new language later in life, which is much harder than picking up a new language at a young age. Early customers of Nova Forge include Reddit, Sony, and Booking.com.