Amazon Unveils Nova: A New Era of Multimodal AI Models
Nova: Amazons neue multimodale KI-Modell-Familie ist da 🔗
Amazon has introduced a new family of multimodal AI models called Nova at its re:Invent event. Nova includes four versions: Micro, Lite, Pro, and Premier, each designed for different tasks and capabilities. The models can handle text, images, and videos, with Micro focusing solely on text for low latency, while Lite processes text, images, and videos but with less accuracy. Pro strikes a balance between speed, cost, and accuracy, and Premier is the most powerful and expensive model, capable of complex tasks. All versions support multiple languages, primarily English, and are available to Amazon customers through AWS Bedrock for fine-tuning. Additionally, Nova includes tools for generating and editing images (Nova Canvas) and short videos (Nova Reel).
What are the different versions of the Nova AI models?
The Nova AI models come in four versions: Micro, Lite, Pro, and Premier, each designed for specific functionalities and performance levels.
What can Nova Canvas and Nova Reel do?
Nova Canvas is a tool for generating and editing images, while Nova Reel allows users to create and edit short videos of up to six seconds long.
How does Amazon ensure the security of the Nova models?
Amazon promises that all Nova models come with high-security measures, although details about the training data and model structure are kept confidential.