December 4 News.Amazontoday announced the launch of a new series of AI-based models branded as "Nova" and will be available through AWS' Amazon Bedrock model library.
Amazon said in a blog post that there are currently three "understanding" models to choose from:
- Amazon Nova Micro: a text model for "speed and cost optimization".
- Amazon Nova Lite: a "very low-cost" multimodal model that inputs images, videos, and text to generate text.
- Amazon Nova Pro: a "powerful" multimodal model.
1AI notes that the company is also training a model called Amazon Nova Premier, theIt is said to be "the most powerful multimodal model for complex reasoning tasks."Amazon aims to offer Nova Premier "early 2025". Amazon aims to offer Nova Premier "by early 2025".
Amazon has also released a content generation model:Amazon Nova Canvas (image generation model) and Amazon Nova Reel (video generation model)The company said the models have "watermarking capabilities" to "promote responsible AI use". The company says the models have "watermarking capabilities" to "promote responsible AI use".
In addition, Amazon plans to release voice-to-speech and "native multimodal-to-multimodal" models later in 2025.
Amazon announced these new models at the AWS re:Invent conference currently underway in Las Vegas. At the show, the company also said that it is working with Anthropic (in which Amazon has invested $8 billion) to build a massive AI compute cluster that relies on its Trainium 2 chip. "When complete, it promises to be the world's largest AI compute cluster to date for Anthropic to build and deploy its future models," Amazon said.
Amazon is also working on a redesigned, AI-powered Alexa, but while the voice assistant was initially planned to launch this fall, its reportedly been pushed back to next year.