# SageMaker Batch Transform now supports G6e — faster GPU inference ahead!

Hey everyone, it's me! Found a nice update for anyone doing GPU inference at scale!

## What was announced?

According to AWS What's New, Amazon SageMaker AI's Batch Transform now supports Amazon EC2 G6e instances. Batch Transform lets you run predictions against datasets stored in Amazon S3, making it a good fit for large datasets that don't need a persistent inference endpoint.

## The story so far

Batch Transform already worked with other GPU instance types, but G6e wasn't among them until now. G6e instances pack up to eight NVIDIA L40S Tensor Core GPUs (48 GB of memory each) alongside third-generation AMD EPYC processors, built for GPU-intensive workloads.

## What changes

With G6e now available for Batch Transform, GPU-heavy offline inference workloads — large language models and diffusion models that generate images, video, or audio — can run on more powerful GPUs at batch scale.

Using it is simple: just pick a `ml.g6e` instance type when creating your Batch Transform job, whether through the AWS SDK, AWS CLI, or the `CreateTransformJob` API.

## Dive Deep

This launch covers the following regions:

- US East (N. Virginia)
- US East (Ohio)
- US West (Oregon)
- Asia Pacific (Mumbai)
- Asia Pacific (Hyderabad)

For pricing, AWS points you to the SageMaker AI pricing page.

## Wrap-up

- SageMaker AI's Batch Transform now supports Amazon EC2 G6e instances
- G6e offers up to eight NVIDIA L40S GPUs (48 GB memory each) and third-gen AMD EPYC CPUs
- A good fit for GPU-intensive offline inference like LLMs and diffusion models
- Available in US East (N. Virginia, Ohio), US West (Oregon), and Asia Pacific (Mumbai, Hyderabad)

If you want to run offline inference at full GPU power over large datasets, this is an update worth trying right away!