Amazon Nova 2.0 Omni (low) is a unified multimodal foundation model released in December 2025 as part of the Nova 2.0 family. It is built on an "any-to-any" architecture that natively processes inputs across text, images, video, and audio while generating both text and image outputs. The low designation refers to the model's reasoning depth setting under Amazon's "extended thinking" framework, which allows developers to adjust the depth of systematic reasoning to optimize for latency and cost.
The model supports a substantial context window of up to 1 million tokens, enabling it to analyze and reason over extensive data sources such as long-form videos, large codebases, and hundred-page documents in a single prompt. In its low-reasoning configuration, the model is optimized for high output speed and lower time-to-first-token (TTFT), making it suitable for real-time applications and high-volume tasks that require multimodal understanding without the overhead of deep reasoning chains.
As a proprietary model integrated into the Amazon Bedrock ecosystem, it includes built-in safety guardrails and supports enterprise customization. Its ability to handle cross-modal retrieval and generation within a single model is designed to simplify agentic workflows and complex content creation tasks, such as generating marketing campaigns from mixed-media assets or performing automated video metadata generation.