Local models with Ollama

Run everything on your own machine: Ollama serves the model, Dolphin Pod talks to it, and nothing leaves your computer.

Install Ollama from https://ollama.com and start it.
Pull a thinking model, for example:
```
ollama pull qwen3.5:2b
```
Larger options: qwen3.5:9b, gpt-oss:20b (supports low / medium / high effort), deepseek-r1:14b.
In Dolphin Pod, add the Ollama provider (default http://localhost:11434) and choose the model for your agents.

Models are checked with Ollama’s own capability list; only models that support thinking are selectable.