At its AWS re:Invent conference, Amazon unveiled a new lineup of advanced AI language models under the Nova brand. These models will be available to AWS cloud customers through the Amazon Bedrock model library.
Currently, three “comprehension-focused” AI models are accessible:
- Amazon Nova Micro: A text-based model optimized for speed and cost-efficiency.
- Amazon Nova Lite: A budget-friendly multimodal model that generates text based on inputs like text, images, or videos.
- Amazon Nova Pro: The most advanced multimodal model in the lineup.
Additionally, Amazon Nova Premier is still in training. Once complete, it aims to handle complex reasoning and problem-solving tasks.
The company also showcased two creative AI systems:
- Amazon Nova Canvas, designed for generating images.
- Amazon Nova Reel, focused on creating videos.
To promote ethical AI use, content generated by these models will include metadata tags for transparency.
In 2025, Amazon plans to release two more models: one designed for voice interaction and another tailored to handle complex multimodal queries and responses.
Partnership with Anthropic and Expanded AI Capabilities
During the event, Amazon announced a collaboration with Anthropic, building a massive AI compute cluster powered by Amazon’s Trainium 2 accelerators. Anthropic, which received an $8 billion investment from Amazon, will use this cluster to develop and deploy its AI projects.
Interestingly, Apple also participated in the conference, revealing it uses Amazon-designed chips for AI development.
Amazon continues to refine its Alexa voice assistant, although its updated version, originally expected last fall, is now likely delayed until next year.
