is there a list of llm models that are able to handle api calls and responses? groq, deepseek, mistral just appear to freeze, wheras openai o4 manages to work