Models

Local models

Connect a local inference engine and keep model execution on your computer.

Connect an engine

Open Settings → Local engine, enter the compatible server address, and test the connection. Available models are read from the engine after a successful test.

Copy-ready example http://127.0.0.1:1234/v1

Getting the best results

Local models differ in their ability to follow multi-step requests and work with project tools. Choose a capable instruction model for complex repository edits and longer tasks.

Choose a capable model

For repository editing, use an instruction-tuned model with strong coding and tool-use capabilities.