AI model · by Amazon
Nova Premier
Nova Premier is Amazon's top-performing multimodal model, designed for intricate work involving text, images, video, documents, and code.
- 1M tokens context
- Closed weights
- Checked against Amazon's page 19 Oct 2026
What Nova Premier is for
Amazon says Nova Premier is its strongest multimodal model for demanding tasks, capable of handling text, images, video, documents, and code. Amazon also says that getting started with Amazon Nova requires reaching the models via an API through Amazon Bedrock.
What Amazon says it does well
- Comprehends text, images, video, documents, and code as inputs
- Context window of 1M tokens
- Supports 200+ languages, optimized for 15 languages including English, German, Spanish, French, Italian, Japanese
Limits Amazon mentions
- Output is text only
- Does not support fine tuning or provisioned throughput
- Bedrock Guardrails and Bedrock Evaluations support text only
- Max output tokens is 10K