How to run a model
Open a model card to see the available options. The setup depends on the source you choose.
- Ollama. In codef, open Models and downloads → Installed → Install by Ollama name. Paste the model name from the card.
- GGUF. Open Models and downloads → Hugging Face, find the repository, and choose a file. Use the GGUF import option if the file is already on your computer.
- API. Connect a cloud provider in Settings → Models and API. Enter your provider key and the model ID from the card.
Need a newer Ollama version?
Quit codef, install and start Ollama, then reopen codef. It will connect to the running server at the default local address.
How much memory do I need?
The download size tells you how much disk space the files take up. Running the model requires extra memory, including space for context. If the model is slow, try a smaller variant or reduce the context in Settings → Models and API.
Choose a variant with tool support if you want the agent to work with files or search the web. Web search contacts an external service. You can turn it off in the agent settings.