Release Note - Oct 29, 2026

General

This release introduces new GPU-enabled AWS G6e and G7 GPU instances. Please note that these instance types currently have limited availability on AWS, so tasks that use them may experience longer wait times before starting.

New AWS Instance Families Include

GPU-Enabled G6e and G7 instances:

G7 instances use NVIDIA RTX PRO 4500 Blackwell Server Edition GPUs, with up to 8 GPUs per instance and 32 GB of memory on each GPU. They are designed for AI inference, graphics, and data analytics, and AWS reports up to 4.6x faster AI inference than the previous-generation G6 instances.

G6e instances use NVIDIA L40S Tensor Core GPUs, with up to 8 GPUs per instance and 48 GB of memory on each GPU, which is double the GPU memory of G6 instances. They are well suited to cost-efficient generative AI inference and to fine-tuning or training smaller models on a single node, and AWS reports up to 2.5x better performance than G5 instances.

Limited availability: G6e and G7 instances are in high demand and have low availability on AWS. Depending on current AWS capacity, tasks configured to use these instances may wait longer to be provisioned or may not start right away. For time-sensitive analyses, we recommend using a more widely available GPU instance type where your workflow allows it.

Resources

To learn more about the newly introduces instance families, visit AWS pages for G6e instances and G7 instances. For the full list of AWS instances available on the platform, see Amazon Web Services.

If you have questions about supported instances, please contact your Velsera/Seven Bridges representative or reach out to the support team at [email protected].