Skip to content

Amazon Nova Model Family Overview

Amazon Nova is Amazon's proprietary suite of foundation models (FMs) hosted on Amazon Bedrock, categorized across Understanding, Creative, and Speech capabilities.

1. Understanding Models

(Process text, image, and video inputs to deliver reasoning, extraction, and generation)

  • Amazon Nova Premier
  • Core Capability: The most capable multimodal model in the Nova family.
  • Primary Use Case: Complex multi-step reasoning tasks and serving as the top Teacher Model for distilling custom, lightweight models.
  • Amazon Nova Pro
  • Core Capability: Highly capable, balanced multimodal model.
  • Primary Use Case: Offers the best combination of accuracy, speed, and cost for a wide range of enterprise tasks.
  • Amazon Nova Lite
  • Core Capability: Very low-cost, lightning-fast multimodal model.
  • Primary Use Case: Rapid processing of image, video, and text inputs for high-volume, interactive applications.
  • Amazon Nova Micro
  • Core Capability: Text-only model with ultra-fast generation speeds.
  • Primary Use Case: Delivers the lowest latency responses at very low cost for simple text-based tasks and real-time chat.

Micro -> Lite -> Pro -> Premier (increasing capability, cost, and latency).

2. Creative Models

(Generate visual media assets from text and image prompts)

  • Amazon Nova Canvas
  • Core Capability: State-of-the-art image generation model.
  • Primary Use Case: Creating high-resolution marketing images, inpainting, color adjustments, and visual style customization.
  • Amazon Nova Reel
  • Core Capability: State-of-the-art video generation model.
  • Primary Use Case: Generating short, professional-grade promotional video clips from text or image prompts.

3. Speech Models

(Real-time, voice-first conversational interactions)

  • Amazon Nova Sonic
  • Core Capability: Speech understanding and voice generation model.
  • Primary Use Case: Real-time, fluid conversational speech interactions and voice automation across multiple languages.