Amazon Nova Lite is a multimodal foundation model designed for low-latency processing and cost efficiency. Developed by Amazon as part of the Nova family, it supports text, image, and video inputs to generate text-based outputs. The model is engineered for high-throughput applications such as real-time customer interactions, document analysis, and visual reasoning. It features a 300,000-token context window, which allows it to process approximately 30 minutes of video or multiple long-form documents within a single request. Its architecture is based on a latent multimodal transformer designed to integrate diverse data modalities for complex visual and linguistic tasks. Nova Lite supports more than 200 languages and includes capabilities for function calling and tool use in agentic workflows. It also incorporates built-in safety guardrails to mitigate the generation of harmful content and address grounding risks.
