Connect an AI Model

Pairwise LM connects to a cloud provider via a standard API endpoint, a model running locally on your PC, or the built-in model. You choose the endpoint, verify it works, and then run your evaluation scripts with full control over every action.

Choose Your Model Source

Cloud API Endpoint

Connect to a cloud AI provider using a standard API endpoint. This option sends your prompts to a remote server, so the text leaves your machine. You can use any provider that supports the standard interface, and you control which endpoint receives your data. Use this option for large models that do not fit on your hardware.

Local PC Model

Run a model directly on your Windows desktop using local hardware. No data leaves your machine when you use this option, ensuring your prompts stay on your computer. You can monitor your GPU memory and speed using your own vendor tools while the agent runs your evaluation scripts.

Built-in Model

The built-in model is ready to use without your own endpoint or API key.

Verify Your Connection

Click the Test button to check that the endpoint answers. It shows response time and the context window size. This is not a substitute for your own benchmark.

Select the Right Model

Use run_on_model to send a hard step to a bigger model. If that model is a cloud endpoint, that step's text goes to the provider.

Quality varies between models, so always review the diff before accepting changes. Use Ask first to pause before every action that changes something. This ensures you see the exact modifications the model makes to your evaluation scripts or result files before they are applied permanently.