Skip to main content
Valkyrie is Azumo’s AI platform for running open language models. You pick a model, deploy it, and call it through a normal OpenAI-compatible API. If you need it to be better at your own task, you can fine-tune it on your data first. Either way, the GPUs are rented for you on demand and shut down when they go idle, so you only pay while a model is actually working.

Quickstart

Go from sign-up to your first API call in a few minutes.

Core concepts

Deployments, fine-tuning, the wallet, aliases, and the MCP server.

Connect a model

Point Claude Code, Cursor, or any OpenAI-compatible client at your model.

MCP server

Let an AI agent run Valkyrie for you.

​
What you can do

  • Run an open model without setting up a single GPU.
  • Fine-tune a model on your own examples so it answers the way you need.
  • Serve any model behind a stable API and a friendly URL.
  • Drive the whole thing from an AI agent through the MCP server.
  • Pay only for what you use, from a prepaid balance. No seats, no idle GPU bills.

​
Who Valkyrie is for

Run a model for your own use

You want to use a capable open model and would rather not rent GPUs, install serving software, or babysit infrastructure. Deploy one and start calling it.

Get better accuracy with fine-tuning

A general model gets you part of the way. Fine-tune it on your own examples and it gets noticeably better at your specific task.

Keep your models private

For companies that want their own models running on dedicated instances, with data and access under their control rather than sent to a third-party API.
Developers and AI agents fit here too: anything that speaks the OpenAI or Anthropic API can call a Valkyrie model, and agents can operate the platform through the MCP server.
New here? Start with the Quickstart, then read Core concepts to see how the pieces fit together.