Changelog #85 - A100 SXM available in North America, Gemma 3, Phi-4 Multimodal, and Mistral Small 3.1 One-Click Models, and more

[David]

Hello, and welcome to this week’s changelog update! Let’s dive into what’s new:

1. A100 SXM available in North America

   You can now deploy on Nvidia A100 SXM Instances in the Iowa region on Koyeb. This means more high-performance GPU options for your workloads in North America.

   

2. New One-Click Models: Gemma 3, Phi-4 Multimodal, and Mistral Small 3.1

   You can now deploy the latest Gemma 3 27B, Phi-4 Multimodal Instruct, and Mistral Small 3.1 Instruct models with one click from the catalog.

   

3. Custom Scale to Zero Idle Timeout for Pro Plans and Above

   For Pro Plan users and above, you can now customize the idle timeout before your Services sleep when Scale to Zero is enabled. Previously, this time was fixed at 5 minutes.

   

4. Best Open Source Multimodal Vision Models in 2025

   Curious about the best open-source multimodal vision models being used today? Check out our blog post where we explore top models including Qwen 2.5 VL 72B Instruct, Pixtral 12B, Phi-4 Multimodal Instruct, and more.