SageMaker AI introduces serverless model customization for Qwen3.6

Amazon SageMaker AI has expanded its capabilities to include serverless model customization for the Qwen3.6 model, which features 27 billion parameters. This enhancement utilizes supervised fine-tuning (SFT) and reinforcement fine-tuning (RFT) methods. The Qwen3.6 model is part of a widely recognized open-weight model family developed by Alibaba Cloud. This new feature builds upon SageMaker’s existing support for fine-tuning the Qwen3.5 model and other popular models.

Previously, users could deploy the Qwen3.6 base model on SageMaker AI. With this update, users can now customize the model to fit specific domains and workflows. This customization allows for the adaptation of foundational models with proprietary data, ensuring they better reflect domain-specific knowledge, terminology, and quality standards. Fine-tuning offers the advantage of starting with a robust base model and tailoring it to meet specific use cases, such as enhancing accuracy for domain-specific tasks, aligning outputs with organizational tone, or boosting performance on new tasks using labeled data.

By offering serverless customization, SageMaker AI manages all necessary infrastructure provisioning and training orchestration. This means users can concentrate on data and evaluation without the need to handle cluster management, paying only for the resources they utilize.

The serverless model customization for Qwen3.6 is available in several regions, including US East (N. Virginia), US West (Oregon), Asia Pacific (Tokyo), and EU (Ireland). To begin, users can navigate to the Models page within Amazon SageMaker Studio to initiate a customization job or utilize the SageMaker Python SDK for programmatic access. Additional information can be found in the Amazon SageMaker AI model customization documentation.