Amazon: Nova 2 Lite

Amazon Nova 2 Lite sits within the Nova family as the lightweight tier, positioned for fast responses and economical deployment when richer Nova variants would be excessive for the task. The catalog lists text, image, video, and PDF inputs paired with text output, along with attachments, reasoning, tool calling, and temperature control, suggesting it is meant to function as a versatile general assistant rather than a narrowly scoped utility. Its very large context window relative to the per-token price makes it a reasonable candidate for analyzing lengthy documents, transcripts, or mixed-media prompts where cost accumulation is a real concern.

AWS has framed the model around practical enterprise workflows, and a follow-up AWS Machine Learning blog post specifically demonstrates object detection use cases built on top of it, indicating that vision understanding is intended to go beyond simple captioning and into applied perception tasks. The combination of multimodal inputs, reasoning capability, and tool calling supports agent-style patterns where the model reads attachments, reasons about their contents, and orchestrates downstream actions. For teams already operating inside AWS, the Kilo Gateway routing keeps adoption low-friction while letting them experiment with a lighter reasoning model before committing to higher-cost tiers.