Amazon has launched its own AI model named Nova. It is a state-of-the-art foundational model designed to facilitate generative AI tasks. This marks a bold venture that Amazon has tapped into in the field of generative AI, in addition to its association with the Anthropic. 

What does Amazon Nova do?  

Think of Foundational AI models that can help the users analyze documents, make videos, generate images from text, and finally generate marketing content, Amazon Nova comes to the rescue. It is a foundational AI model built to serve purposes namely, Understanding and Creative Generation. 

How does it Work?

The Understanding Model accepts text, image, and video inputs for analysis and text generation, while the Creative Content Generation Model accepts text and image input for image or video output. This positions Amazon on par with the top players in the Gen-AI market.

What are the 3 models under Amazon Nova?  

The current offering includes 3 types of models, while preparations are en route to launch the 4th one by early next year. The three models are: 

1. Amazon Nova Micro- It is a text-only model that specializes in text summarization, translation, content classification, interactive chat and brainstorming, simple mathematical reasoning, and coding. It delivers the lowest latency responses.
2. Amazon Nova Lite: It is a multimodal mode, compatible with all modes- image, text, and video. It specializes in real-time customer interactions, document analysis, and visual question-answering tasks with high accuracy. It can analyze up to 30-minute videos in a single request.
3. Amazon Nova Pro: It is a highly capable multimodal model specializing in processing both visual and textual information and excels at analyzing financial documents. With an input context of 300K tokens, it can process code bases with over fifteen thousand lines of code.

Highly Customizable 

The foundational model developed by Amazon offers utter customization capabilities making it an attractive choice for enterprises. The models can be trained with text, images, and video to understand industry-specific terminology and optimize its services for their use.

Availability 

These models are exclusively available on Amazon Bedrock in the US East (N. Virginia) AWS region. Amazon Nova Micro, Lite, and Pro are also available in the US West (Oregon), and US East (Ohio) regions via cross-Region inference. Additionally, these are at least 75% cheaper than the best-performing models in their respective intelligence classes in the Amazon Bedrock.  

Quotient of Understanding

The Amazon Nova model learns what matters most to customers by analyzing their data, including text, images, and videos. Then, Amazon Bedrock trains a private, fine-tuned model to deliver personalized responses.

These models understand and generate content in over 200 languages, with particularly strong capabilities in English, German, Spanish, French, Italian, Japanese, Korean, Arabic, Simplified Chinese, Russian, Hindi, Portuguese, Dutch, Turkish, and Hebrew. 

This marks a new milestone in the Gen-AI industry as Amazon which already has an AWS user base launches its own Gen-AI model.