Amazon SageMaker AI Gets Flexible Sequence Lengths for Fine-Tuning

Amazon SageMaker AI now lets you customize sequence length parameters when fine-tuning open-weight models like Llama, Gemma, and Qwen—no more stuck with fixed defaults. You can now set sequence lengths up to 16K tokens for SFT, DPO, and RFT jobs, giving you the flexibility to handle everything from short Q&A pairs to multi-page documents. The feature is live in US East (N. Virginia), US West (Oregon), Asia Pacific (Tokyo), and Europe (Ireland).