Released by Amazon in December 2025, Nova 2 Omni is a unified multimodal foundation model capable of processing inputs across text, images, video, and audio while natively generating both text and image outputs. It is part of the Nova 2 family, designed to eliminate the complexity of coordinating multiple specialized models for multimodal tasks. The model supports a 1-million-token context window, allowing it to ingest and analyze massive datasets, such as multi-hour videos or extensive technical documentation, in a single prompt.
The Non-reasoning designation refers to the model's standard operating mode, distinguishing it from the "extended thinking" or reasoning mode available in the same family. While the reasoning mode uses configurable step-by-step logic to solve complex planning and agentic tasks, the non-reasoning profile is optimized for low-latency and cost-effective execution of general tasks, including document summarization, transcription, and image generation.
Nova 2 Omni features native support for image editing via natural language and provides high-fidelity speech transcription and translation. It includes integrated safety guardrails and is designed for enterprise-level deployment, supporting over 200 languages for text processing and 10 languages for speech input.