We're back with a new round of updates here at Vast.ai! These past few months, we've expanded our offerings so you can get the most out of your time and budget – with new hardware options, fresh templates, and continued enhancements to elevate your experience.

To that end, we're pleased to offer NVIDIA B200 GPUs for rent! These incredibly powerful machines deliver exceptional performance for develop-to-deploy pipelines and all of your most demanding enterprise AI tasks.

But that's not all. Here are the details on what's new!

NVIDIA Cloud GPU Updates

As mentioned, the NVIDIA B200 joins our growing GPU lineup, bringing you more power and options than ever.

Featuring 192 GB of HBM3e memory with extremely high bandwidth, the B200 is ideal for high-throughput AI workloads like fine-tuning, Retrieval-Augmented Generation (RAG), and LLM inference. For additional information and pricing details, learn more here.

Fresh New Templates

We've also rolled out several templates that expand what you can build and run on Vast. Jump-start your creative and technical projects with templates such as:

• InvokeAI – A creative engine providing a professional web interface for AI image generation using Stable Diffusion models.

• Langflow + Ollama – A visual AI workflow platform with an intuitive drag-and-drop interface in Langflow and a hosted Ollama API server with models like Llama, Mistral, and more.

• Ostris AI Toolkit – A complete environment for training custom Stable Diffusion models, LoRAs, and other AI image generation models using your own datasets.

• ACE Step – An AI-powered music generation system that lets you create original music through a simple web interface.

Other Improvements

We've recently updated our Discord server to make conversations more focused, relevant, and helpful for you. Didn't know we had a Discord? Join us there today for direct access to updates, technical guidance, and our supportive user community!

We've also made an important update to how API keys are handled. To strengthen security, API keys are now one-time viewable and not readable anywhere on Vast. Be sure to store your key off of Vast when it's created, as it won't be visible again afterward. Vast.ai API endpoints will no longer return API keys as part of their response (see details in change log below).

That wraps up the updates for October 2025!

Our Commitment

From the new B200 GPUs to convenient templates that help you move from idea to execution, this month's updates continue our mission to make high-performance compute accessible to everyone. We're here to help you build faster and create more, with a platform that grows with your needs.

Need assistance? Reach out to our customer support team at support@vast.ai or join our Discord server anytime for tips, troubleshooting, and community insights.

Change Log

InvokeAI Template now available!

InvokeAI is a leading creative engine for Stable Diffusion models that provides a professional web interface for AI image generation. It's perfect for artists, designers, content creators, or anyone wanting to generate stunning AI artwork. Find the InvokeAI template on Vast.ai here.

Langflow Template & Guide are now live!

This template gives you a complete visual AI workflow platform with Langflow and a hosted Ollama API server running in Docker containers. Langflow provides an intuitive drag-and-drop interface for building complex AI applications, while Ollama serves as your private LLM backend with models like Llama, Mistral, and more. Learn how to launch Langflow on Vast.ai here.

Ostris AI Toolkit Template now available!

This template gives you a complete Stable Diffusion training environment with the Ostris AI Toolkit running in a Docker container. The toolkit provides an intuitive web interface for training custom Stable Diffusion models, LoRAs, and other AI image generation models with your own datasets. Find the Ostris template on Vast.ai here.

ACE Step Template now available!

This template gives you a hosted ACE Step music generation server running in a Docker container. ACE Step is an AI-powered music generation system that lets you create original music through a web interface. It's perfect for musicians, content creators, developers, or anyone who wants to experiment with AI music generation. Find the ACE-Step template on Vast.ai here.

Vast Discord Update: Unlock Your Channels!

We’ve streamlined our Discord to make it cleaner, faster, and more relevant to you. Here’s what’s changed.

Connect your Vast account to instantly unlock channels designed for your role:

• Hosts → Access to #Hosting, #Hosting-rigs, #Support-hosting. (More host channels coming soon!)
• Clients → Unlock support rooms for templates, feature requests, and more.

API Key Security Update + API Breaking Changes

API keys are now one-time viewable and not readable anywhere on Vast. You will need to store your API key off of Vast at the moment it is created, as it will only be visible this one time. If you need access to your existing API key, and did not store it off of Vast, you will need to delete it and create a new one. Vast.ai API endpoints will no longer return API keys as part of their response. However, we've added key_id to responses specifically for cases like this. The newly added key_id can be used to verify a user instead of api_key in certain scenarios. The key_id associated with a user's API key is returned from the show api keys endpoint / command.

key_id is now available for these API calls:

• create_user - POST /api/v0/users
• log in (regular and oauth and tfa) - PUT /api/v0/users/current
• update_user - PUT /api/v0/users
• get_current_user - GET /api/v0/users/current
• get_root_user - GET /api/v0/users/root
• connect_discord - GET /api/v0/oauth/connect-discord/
• disconnect_discord - DELETE /api/v0/admin/disconnect-discord/
• get_subaccounts - GET /api/v0/subaccounts/

The following API key endpoints are now deprecated:

• GET /api/v0/auth/apikeys/{id}/
• PUT /api/v0/commands/reset_apikey/

And the following will now have null for the key values:

• GET /api/v0/auth/apikeys/
• GET /api/v0/team/apikeys/

New NVIDIA B200s added!

B200 GPUs are ideal for high-throughput AI workloads like fine-tuning, Retrieval-Augmented Generation, and LLM inference. These GPUs feature 192 GB of HBM3e memory with extremely high bandwidth.

• Learn more
• Rent now