Hello, and welcome to this week’s changelog update! Let’s dive into what’s new:
-
Scale to Zero in Public Preview
Scale to Zero is now in public preview! With this feature enabled, services with no traffic automatically go to sleep and wake automatically on request. It’s perfect for optimizing infrastructure efficiency, especially for GPU-based inference workloads.
-
A100, L40s, and L4 price drops
We’ve reduced the cost of three of our GPUs, making high-performance computing more accessible:
- A100: from $2.70 to $2/hour
- L40S: from $2 to $1.55/hour
- L4: from $1 to $0.70/hour
-
Enhanced experience with service mesh
The service mesh is now ready as soon as your app starts. We’ve resolved an issue leading to delays in mesh availability. This caused deployment failures for services requiring private networking at startup.
-
Enhanced experience with Volumes created from Snapshots
Volumes created from Snapshots now display the Snapshot name, so you can easily recall their source. Previously, you had to recall whether a Snapshot was used to create a Volume.