right now it's qwen, gpt-oss and ollama

Yup users will be able to switch models