SageMaker AI Serverless Customization Adds Muse Glimmer, Qwen3.8, and Ministral 3!

Hi everyone, it's Shii-chan! SageMaker AI can customize more models now!

AWS What's New aws.amazon.com

What was announced?

AWS What's New announced that SageMaker AI serverless model customization now supports these models:

• Muse Glimmer
• Qwen3.8 27B
• Ministral 3 in 3B, 8B, and 14B

In addition to deploying them, you can now adapt them to your domains and workflows.

What changes

Techniques available:

• SFT: improve accuracy on domain-specific tasks with labeled data
• DPO: align outputs with your organization's tone using preference data
• RFT: improve performance on new tasks using reward signals

All work with either LoRA or full fine-tuning. The idea is that a smaller, right-sized model tuned on your data can match frontier-model quality on your tasks while cutting cost and latency.

Dive Deep

With serverless customization, SageMaker AI handles infrastructure provisioning and training orchestration, so you focus on data and evaluation. You pay only for what you use.

Model families already supported:

• Nova
• Gemma 4
• Nemotron 3.5
• Qwen
• Llama
• gpt-oss
• DeepSeek

Available Regions:

• US East (N. Virginia)
• US West (Oregon)
• Asia Pacific (Tokyo)
• Europe (Ireland)

Wrap-up

• Muse Glimmer, Qwen3.8 27B, and Ministral 3 (3B, 8B, 14B) are supported
• Choose SFT, DPO, or RFT with LoRA or full fine-tuning
• No infrastructure to manage, pay as you go
• Available in US East, US West, Tokyo, and Ireland

This one is for teams who want to tune models on their own data without touching GPU clusters!