Learn which LLMs Thunders uses, where they operate, and how your data is processed, all explained in one clear overview.
1. Where are the LLMs running?
Thunders runs its AI models in Azure AI Foundry, a managed cloud infrastructure hosted in Microsoft Azure datacenters in Europe and fully dedicated to our platform.
For some specific actions, Thunders uses Gemini running on European datacenter of Google, and fully dedicated to our platform
No customer data is stored outside these runtime environments.
2. Which LLMs are used?
We use several state-of-the-art LLMs to process natural-language instructions:
Azure Open AI
Azure Anthropic
Gemini
Note: These models are not trained on customer data.
3. What data do LLMs process?
LLMs only process the information required to execute a test step, such as instructions like *"Click Continue"* or *"Enter email address"*.
Sensitive information (credentials, files, proprietary content) is never stored.
Data is transient and used only during execution.
4. Key points for security and privacy
Customer data is not retained.
Models run in a secure, controlled environment.
The setup complies with industry best practices for data protection.
5. Self-hosted: use your own models
Everything above describes the Thunders cloud. If you run Thunders self-hosted, the AI layer is yours to choose: you connect your own model endpoints, under your own contracts and your own keys, and no test data leaves the infrastructure you control.
You configure three model roles, so you can match each one to the cost and capability you want:
A fast model for short, high-volume decisions
A full model for the heavier reasoning work
A visual model for the steps that need to look at the screen
You can point all three at one model, or mix them. A common setup is a small, inexpensive model for the fast role and a stronger one for the full role.
Thunders connects to model endpoints of these kinds:
Azure AI Foundry, including models you have deployed in your own Azure subscription
OVHcloud AI Endpoints, for teams that want their models hosted in Europe under a European provider
Google Gemini, by API key
Google Vertex AI, in your own Google Cloud project
Any OpenAI-compatible provider. Give Thunders the endpoint URL, an API key and a model name, and it works. This covers most hosted model services and most self-hosted model servers, including open-weight models you run on your own hardware.
This is the deployment to choose when your policy requires that model traffic stay inside a boundary you own, or when you already have a model contract you want to keep using.
Learn more: Thunders Hybrid Deployment Guide
Summary:
On the Thunders cloud, our LLMs convert human-readable test instructions into precise actions while ensuring secure, private, and reliable execution. Self-hosted, you bring your own models and keep model traffic inside your own boundary.
Contact Us
If you have any questions about our security practices don't hesitate to check this article.
For further details, please feel free to reach out through the chat widget at the bottom of your screen or on [email protected]