SageMaker AI supports serverless customization for Gemma 4 models
Amazon SageMaker AI now offers serverless model customization for Google DeepMind's Gemma 4 E4B and 31B models, including supervised fine-tuning, direct preference optimization, and reinforcement fine-tuning. This allows users to tailor foundation models with proprietary data for domain-specific tasks without managing infrastructure. Serverless customization is now available for Gemma and expands options for other families like Nova, Nemotron 3, Qwen, Llama, gpt-oss, and DeepSeek.
- →Serverless model customization for Gemma 4 models on SageMaker AI
- →Expanded serverless customization options for foundation models
- →Benefits and availability of serverless model customization
Features (1) ›
- Serverless model customization for Gemma 4 models on SageMaker AI
Amazon SageMaker AI now supports serverless customization for Google DeepMind's Gemma 4 E4B and 31B models using SFT, DPO, and RFT. This feature allows users to adapt foundation models with proprietary data for specific domains and workflows.
Enhancements (1) ›
- Expanded serverless customization options for foundation models
The availability of Gemma 4 models for serverless customization extends the variety of models that can be adapted on SageMaker AI. This includes models from the Nova, Nemotron 3, Qwen, Llama, gpt-oss, and DeepSeek families.
Notes (1) ›
- Benefits and availability of serverless model customization
Serverless customization allows tailoring foundation models with proprietary data to improve accuracy and performance, with SageMaker AI handling infrastructure provisioning and orchestration. This feature is available in specific AWS regions, and users can access it via SageMaker Studio or the SageMaker Python SDK.
https://aws.amazon.com/about-aws/whats-new/2026/06/amazon-sagemaker-ai-gemma-4/