SageMaker full-tunes 25+ models without servers

Amazon SageMaker now offers serverless full fine-tuning for more than 25 open-source models, letting teams update every parameter in GPT-OSS, Gemma, Llama, Nemotron, and Qwen families without provisioning training infrastructure, according to AWS’s launch notice. The service is initially available in four regions and charges only for use; its model catalog includes weights from 0.6B to 120B parameters. For operators following the open-weight market’s expanding leverage, the fact to file is that deep customization is becoming a managed-cloud job rather than a cluster-engineering project.