Amazon just made its biggest AI play yet. At AWS re:Invent, the company unveiled four Nova 2 models spanning reasoning, speech, and multimodal capabilities, plus two new services that let enterprises build custom AI and deploy browser agents at scale.
The Nova 2 lineup includes Lite and Pro reasoning models, Sonic for real-time voice AI with a million-token context window, and Omni, an industry-first model that processes text, images, video, and audio while generating both text and images. Pro can handle hours of audio and 750,000 words simultaneously, while both reasoning models pack web search and code execution built in.
Nova Forge introduces "open training," giving organizations access to model checkpoints at every training stage so they can blend proprietary data with Amazon's curated datasets. Reddit's CTO says the approach is replacing multiple content moderation models with a single, more accurate solution. Meanwhile, Nova Act delivers 90% reliability on browser automation workflows, with Hertz reporting 5x faster software delivery after adopting it for QA testing.
The Bottom Line: Amazon is betting that customizable frontier models plus ultra-reliable agents will pull enterprises deeper into the AWS AI stack.
QUICK FACTS
- 4 new models: Nova 2 Lite, Pro, Sonic, Omni
- Nova 2 Sonic supports 1 million token context window
- Nova Act achieves 90% reliability on browser automation
- Hertz accelerated software delivery 5x using Nova Act
- Partners include Reddit, Booking.com, Sony, 1Password
