Run open-weight models in your app, on a subscription.
Two kinds of endpoint. Pick the one you need.
Cloud hosts the model you fine-tuned. App Server hosts ready-to-use open-weight models so you can put one in your app without training anything.
Cloud
Your custom model
You fine-tuned it with the Framework. Cloud trains it on managed GPUs and hosts it at your-name.arkor.app.
App Server
Ready-to-use open-weight
The Gemma 4 family is already running. You get an OpenAI-compatible baseUrl and an API key. No training step.
Under a minute, from the web UI.
Open Arkor in your browser, pick a model from the Gemma 4 family, copy the OpenAI-compatible baseUrl. The Framework npm package is not required. Start free on the Hobby plan.
More from the toolkit.
Code-first SDK for fine-tuning open-weight models.
Learn moreInspect and manage training runs from your browser or coding agent.
Learn moreTrain custom models on managed GPUs and use them instantly.
Learn moreDistill a large model into a smaller one you can serve cheaply.
Learn more