AWS Debuts ‘Most Capable’ Model Yet: Amazon Nova Premier

AWS is unveiling the GA of Amazon Nova Premier, the latest—and most capable—model joining AWS’ existing Amazon Nova understanding models, as phrased by the company. Engineered for complex tasks and as a teacher for model distillation with a context length of one million tokens, Premier delivers advanced capabilities with performance, cost-efficiency, and speed at the forefront.

According to Danilo Poccia, chief evangelist (EMEA) at AWS, “Nova Premier excels at complex tasks that require deep understanding of context, multi-step planning, and precise execution across multiple tools and data sources.” And with its large context window, Premier can process extremely long documents or codebases.

As a teacher model, Premier—in conjunction with Amazon Bedrock Model Distillation—can be utilized to produce unique versions of Nova Pro, Lite, and Micro designed around  specific needs while centering capability, cost-efficiency, and low latency. With this opportunity for customization, using Premier as a teacher enables organizations to create models with Premier’s advancements for smaller, faster deployments.

Furthermore, “Unlike traditional fine-tuning that requires human feedback and labeled examples, with model distillation you can generate high-quality training data by having a teacher model produce the desired outputs, streamlining the data acquisition process,” explained Poccia.

As an example, AWS leveraged Premier to distill Nova Pro for complex tool selection and API calling. This resulted in Nova Pro delivering 20% higher accuracy for API invocations, while matching the performance of the teacher and maintaining Pro’s speed and cost efficiency.

According to AWS, Premier is the fastest and most cost-effective model in Amazon Bedrock at its intelligence tier. It is “comparable to the best non-reasoning models in the industry and is equal or better on approximately half of these benchmarks when compared to other models in the same intelligence tier,” noted Poccia.

“We evaluated Nova Premier on a broad range of benchmarks across text intelligence, visual intelligence, and agentic workflows. Nova Premier is the most capable model in the Nova family as measured across 17 benchmarks,” Poccia added.

Many customers are already experiencing the benefits of Nova Premier. Curtis Allen, senior staff engineer at Slack, remarked how, “Amazon Nova Premier has been outstanding in its ability to execute interactive analysis workflows, while still being faster and nearly half the cost compared to other leading models in our tests.”

Henry Ehrenberg, co-founder of Snorkel AI, commented how, “Accelerating real-world AI deployments—not just prototypes—requires the ability to build models that are specialized for the unique needs of real-world applications. We’re excited to see AWS pushing efficient model customization forward with Amazon Bedrock Model Distillation and Amazon Nova Premier. These new model capabilities have the potential to accelerate our enterprise customers in building production AI applications, including Q&A applications with multimodal data and more.”

To learn more about Amazon Nova Premier, please visit https://aws.amazon.com/.