# SageMaker adds full fine‑tuning to AI Serverless customization

> AWS now lets users run full fine‑tuning of LLMs on SageMaker AI Serverless, removing the need for GPU clusters and boosting accuracy on large datasets.

Oossa · 2026-11-19 · https://oossa.com/en/sagemaker-adds-full-fine-tuning-to-ai-serverless-customization

Amazon SageMaker AI Serverless Model Customization now supports Full Fine‑Tuning (FFT). The service runs FFT jobs without requiring users to provision GPUs or set up clusters. In tests on a 110,000‑example code dataset, FFT reached 90 % semantic correctness, outperforming LoRA on all quality metrics. The same code that launches a LoRA job can be used for FFT by switching a single parameter.

## The facts

- FFT support announced on 2026‑11‑19 ("Full Fine‑Tuning (FFT) on SageMaker AI Serverless Model Customization" – published Thu Nov 19 2026).
- FFT fine‑tuned a Qwen3‑4B model on 110 K code instructions, achieving 90 % LLM‑as‑Judge semantic correctness.

## Why it matters

Enterprises can improve model accuracy on complex tasks without managing GPU infrastructure, lowering operational overhead and cost.

## Sources & references

1. [Full Fine-Tuning (FFT) on SageMaker AI Serverless Model Customization](https://aws.amazon.com/blogs/machine-learning/full-fine-tuning-fft-on-sagemaker-ai-serverless-model-customization/) – AWS, 2026-11-19

Last updated: 2026-11-19
