# Amazon SageMaker adds custom sequence lengths to serverless fine‑tuning

> SageMaker AI now lets users set token limits up to 16K when fine‑tuning open‑weight models in serverless jobs.

Oossa · 2026-10-21 · https://oossa.com/en/amazon-sagemaker-adds-custom-sequence-lengths-to-serverless-fine-tuning

Amazon SageMaker AI announced that its serverless model customization jobs can now accept user‑defined sequence lengths. Previously the jobs ran with a fixed token count. Users can now set dataset_max_len for SFT and DPO jobs, or max_prompt_length and max_response_length for RFT jobs, up to the model’s published maximum—often 16,000 tokens. The feature is live in US East (N. Virginia), US West (Oregon), Asia Pacific (Tokyo) and Europe (Ireland).

## The facts

- Announcement date: 2026-10-21 ("Wed Oct 21 2026")
- Maximum supported length: up to 16 K tokens per model/technique

## Why it matters

Developers can fine‑tune large language models on longer documents without writing custom training code.

## Sources & references

1. [Amazon SageMaker AI now supports custom sequence lengths for serverless model customization jobs](https://aws.amazon.com/about-aws/whats-new/2026/10/sagemaker-custom-sequence-lengths/) – AWS, 2026-10-21

Last updated: 2026-10-21