
Amazon launched Nova Premier, its latest and most innovative AI model to date. The new model comes as part of Amazon's Nova family, which enables users to accomplish intricate tasks through generative AI.
Nova Premier is capable of understanding and generating text, images, and videos. It's now accessible on Amazon Bedrock, the firm's platform where developers create AI tools.
Amazon's Nova Premier model
Nova Premier is for tough tasks that require deep thinking, planning, and precision. Although it doesn't process audio, it has a context length of 1 million tokens, sufficient to comprehend about 750,000 words simultaneously.
While it is slower than Google's Gemini 2.5 Pro in some benchmarks, particularly coding, math, and science, Amazon claims Nova Premier excels in knowledge retrieval and visual understanding.
On SimpleQA and MMMU benchmarks, Nova Premier achieved high scores.
Meanwhile, the cost is competitive, too. It charges $2.50 for 1 million input tokens and $12.50 for 1 million output tokens, as much as Google charges for Gemini 2.5 Pro.
However, Nova Premier is not a reasoning AI model, so it cannot double-check or slow down to provide smarter responses like some competitors.
Amazon suggests using it for AI model distillation instead, training smaller models to do better by emulating Nova Premier's strengths.
Moreover, CEO Andy Jassy stated that the company is developing more than 1,000 generative AI apps.
He further stated that Amazon's AI business is rapidly expanding and now generates billions of dollars annually.