Amazon Nova foundation model
Amazon Nova Lite
Fast, very low-cost multimodal Nova understanding model that accepts text, image, and video input and outputs text.
Key points
- Accepts text, image, and video input and outputs text.
- Provides very low cost and fast responses for multimodal workloads.
- Supports about a 300K token context window.
- Supports more than 200 languages, with optimization for a core language set.
- Supports fine-tuning for supported customization use cases.
When to use it
- Choose Lite for high-volume multimodal tasks that need speed and low cost.
- Use it for image or video understanding when Pro-level reasoning is not required.
- Use through Amazon Bedrock when a lightweight multimodal text-output model fits the task.
Exam tips
- Lite is multimodal; Micro is the text-only Nova understanding model.
- Lite can accept video input but still outputs text, not video.
- Choose Lite over Pro or Premier when low cost and speed are stronger clues than maximum capability.