Nova 2.0 Pro Preview (low) is a specific configuration of Amazon's high-intelligence frontier model released during the AWS re:Invent 2025 conference. It represents the "low" intensity setting of the Adjustable Thinking Budget, a feature that allows developers to toggle step-by-step reasoning on or off and choose between low, medium, and high budget levels to optimize for speed, cost, or reasoning depth. This specific configuration is designed for applications requiring a balance of the Pro model's high intelligence with lower latency for interactive or high-throughput workflows.
As the most intelligent offering in the second-generation Nova family, the model is engineered for complex, multi-step tasks, including advanced software migrations, multi-document analysis, and sophisticated agentic planning. It features a 1-million-token context window, enabling the processing of extensive codebases, hour-long videos, or hundreds of pages of documentation in a single prompt.
Key Capabilities
- Multimodal Reasoning: The model natively understands text, images, video, and speech, allowing for cross-modal analysis and data extraction.
- Agentic Infrastructure: It includes built-in support for tool use, MCP servers, and specialized features like a code interpreter and web grounding for real-time information retrieval.
- Developer Controls: Beyond the thinking budget, the model supports advanced customization through Amazon Nova Forge, which offers early-checkpoint tuning and model distillation options.
- Performance: In its "low" thinking state, the model provides faster response times than the higher reasoning tiers while maintaining significantly higher accuracy and reasoning capabilities compared to the previous generation's Nova Premier model.