AWS

AWS Adds Managed Inference Support for Custom Amazon Nova Models


Executive Summary

Amazon Web Services has announced the general availability of Amazon SageMaker Inference for custom Amazon Nova models. This new capability allows customers to deploy their fine-tuned Nova Micro, Nova Lite, and Nova 2 Lite models on a production-grade, managed service. The offering aims to provide customers with greater control over cost and performance by enabling configuration of instance types, auto-scaling, and other advanced inference parameters for production workloads.

Key Takeaways

* New Service: Custom-trained Amazon Nova models can now be deployed using the managed Amazon SageMaker Inference service.

* Supported Models: The service supports Nova Micro, Nova Lite, and Nova 2 Lite models that have been customized through continued pre-training, supervised fine-tuning, or reinforcement fine-tuning.

* Cost & Performance Optimization: Users can reduce inference costs through optimized GPU utilization on Amazon EC2 G5 and G6 instances, auto-scaling policies, and configurable parameters like context length, concurrency, and batch size.

* Deployment Flexibility: Models can be deployed through the SageMaker Studio UI or programmatically using the SageMaker AI SDK.

* Availability: The service is now generally available in the US East (N. Virginia) and US West (Oregon) AWS Regions.

* Pricing: Customers pay per hour for the compute instances used, with no minimum fees or commitments.

Strategic Importance

This launch creates a seamless, end-to-end AWS-native workflow for customizing and serving Amazon's proprietary foundation models, strengthening the SageMaker ecosystem against competitors and third-party model providers.

Original article