At its reInvent conference on Tuesday, Amazon Web Services (AWS), Amazon’s cloud computing division, announced a new family of multimodal generative AI models it calls Nova.
The four text-generating models are Premier, Pro, Lite, and Micro. Amazon CEO Andy Jassy stated onstage that AWS users may now access Micro, Lite, and Pro on Tuesday, while Premier will be available in early 2025.
Additionally, there is a model called Nova Canvas that generates images and a model called Nova Reel that generates videos. This morning, both were also launched on AWS.
Jassy said,
“We’ve continued to work on our own frontier models,”
“and those frontier models have made a tremendous amount of progress over the last four to five months. And we figured, if we were finding value out of them, you would probably find value out of them.”
The text-generated Nova models can be optimized 1in 5 languages (but primarily English), and are widely variable depending on the sizes and capabilities.
Micro can only take the text and output text. Also, they can deliver the lowering of the bunch latency. This process is a form of texting and also can generate a faster response.
Additionally, the lite edition can process text, video, and image inputs rather quickly. Accuracy, speed risk ranges, and cost are all balanced in pro versions. The premium models are better equipped to handle the intricate workload.
All three versions can analyze the text, images, and videos. All of these are well suited for any of the tasks like digesting documents and the summarization chart, meeting, and chart. The Premier model that AWS is pitching is more of a teaching tool for developing customized customer models than a stand-alone model.
Micro owns more than 128,000 tokens in the context window. That is the meaning of the process and that goes up to around 100,000 words.
Lite and Pro both have 300,000-token context windows. That can be worked out to be around 225,000 words, 15,000 lines according to the computer code, or 30 minutes of footage. More than two million tokens will be supported by the expanded Nova models context windows in early 2025.
Jassy also claims the Nova models are among the fastest in the class. This is among the least expensive places to run. They are also available in the AWS Bedrock and the Amzon’s AI development platforms can be fine-tuned on the text.
Jassy also added
“We’ve optimized these models to work with proprietary systems and APIs, so that you can do multiple orchestrated automatic steps — agent behavior — much more easily with these models,”.
“So I think these are very compelling.”
Read More News
News Source: https://shorturl.at/aniMO
