AWS’ Nova 2 AI Models Launched At re:Invent 2025 As CEO Touts New Innovation

CEO Matt Garman’s keynote at AWS re:Invent 2025 today focused on the launch of four new Nova AI models looking to shake up the AI industry. Here’s what every AWS partner needs to know.

Amazon Web Services launched four new AI models inside its Nova portfolio with the goal of taking AI training, price performance and AI customer opportunity to the next level.

AWS CEO Matt Garman’s keynote at AWS re:Invent 2025 in Las Vegas on Tuesday focused on the $132 billion cloud leader’s new Nova 2 AI models: Lite, Pro, Sonic and Omni.

Garman dubbed Nova 2 Lite as the most cost-effective reasoning model for everyday workloads on the market today.

“Nova 2 Lite compares really favorably in industry benchmarks to models like Claude Haiku 4.5, GPT-5 Mini, and Gemini Flash 2.5,” said Garman on stage in front of thousands of attendees at AWS re:Invent. “It’s our new fast and cost-effective reasoning model suitable for broad set of workloads.”

[Related: AWS Partners ‘Are The Lifeblood,’ Says CEO Garman Ahead Of Channel Program Evolution]

“Nova 2 Lite excels at things like instruction following, tool calling, generating code and extracting information from documents—often matching or exceeding the performance that we see from these comparable models at an industry leading cost performance,” said AWS’ CEO.

Garman said customers can use Nova 2 Lite for automating business process workflow, intelligent document processing and customer support, while also helping software developers with code generation, debugging, refactoring, and migrating systems.

“We think that Amazon Nova 2 Lite is going to be a real work horse, and it’s going to be really popular for a wide variety of use cases out there,” Garman said.

All of the Nova 2 models are now available inside AWS’ AI platform Amazon Bedrock.

AWS CEO Says Nova 2 Pro ‘Really Shines’

AWS’ new Nova 2 Pro AI model is being dubbed Amazon’s most intelligent reasoning model that can process text, images, video, and speech to generate text.

“Pro is our most intelligent reasoning model, and it’s going to be great when you have those really complex workloads,” said AWS’ CEO. “In particular, we look at really important areas where you need your agents to be great, and that’s where Nova 2 Pro really shines—where skills like instruction following and agentic tool use are critical.”

Garman said Nova 2 Pro is ideal for highly complex tasks like agentic coding, long-range planning, and sophisticated problem-solving. He said the model can also serve as a “teacher” for knowledge distillation—transferring its capabilities into smaller, more efficient “student” models for specific domains and use cases.

“If you look at artificial analysis benchmarks in those areas, Nova 2 Pro delivers better absolute results compared to leading models like GPT 5.1, Google Gemini 3 Pro and Claude Sonnett 4.5,” AWS CEO said.

‘You’re Going To Love’ Nova 2 Sonic, Garman Says

Nova 2 Sonic is Amazon’s new speech-to-speech model that unifies text and speech understanding and generation for real-time, human-like conversational AI.

“Sonic offers industry leading conversational quality at awesome price performance. With improved latency and significantly expanded language support, we think you’re going to love it,” Garman said,

Nova 2 Sonic features expanded multilingual support with expressive voices, higher accuracy, and a one-million token context window for sustained interactions, while enabling seamless switching between voice and text.

Nova 2 Omni Is An Industry-First AI Model

Nova 2 Omni is a unified multimodal reasoning and generation model that can process text, images, video, and speech inputs while generating both text and images—an industry first, Garman said.

“Omni is the industry’s first reasoning model that supports text, image, video and audio input, and then it supports text and image generation output,” Garman said.

Nova 2 Omni handles up to 750,000 words, hours of audio, long videos, and hundred-page documents, simultaneously analyzing entire product catalogs, testimonials, brand guidelines, and video libraries at once.

AWS said organizations like Cisco, Siemens, Sumo Logic, and Trellix are already using Nova 2 models for applications ranging from agentic threat detection to video understanding and voice AI assistants.

“So that’s four new industry-leading models for Amazon Nova, and we’re just getting started,” said Garman.

AWS re:Invent 2025 takes place this week in Las Vegas.