Amazon has launched a range of foundational AI models under the Nova brand, now available in the Amazon Bedrock model library on the AWS platform. These models aim to address various user needs, from text understanding to content generation.
Current Nova Models for Understanding
The company’s blog highlights three Nova models in the “understanding” category:
- Amazon Nova Micro: A text-based model designed for speed with minimal resource requirements.
- Amazon Nova Lite: A low-cost multimodal model capable of processing text, images, and video to generate text.
- Amazon Nova Pro: A high-performance multimodal model for more demanding tasks.
Looking ahead, Amazon is developing the Amazon Nova Premier, the most advanced multimodal model for complex logical reasoning. This model is scheduled for release in early 2025.
Content Generation and Future Models
Amazon has also introduced models for content creation:
- Amazon Nova Canvas: Focused on generating images.
- Amazon Nova Reel: Specializes in creating videos, as demonstrated by a sample promotional video for a fictional pasta brand.
Both models feature a watermarking system to encourage responsible AI use.
In 2025, Amazon plans to expand its Nova lineup with two groundbreaking models:
- Speech-to-Speech Model: This model will understand natural language, including tone and intonation, and provide responses that sound natural and engaging.
- Any-to-Any Model: A native multimodal solution capable of processing and generating text, images, audio, and video. This versatile model will allow tasks like format conversion and content editing within a single application.
Broader AI Developments
NIX Solutions reminds that in addition to Nova, Amazon recently increased its investment in Anthropic, a key competitor to OpenAI, showcasing its commitment to advancing AI innovation.
We’ll keep you updated on Amazon’s progress as new models and features become available.