If open ai reasoning models, how to allot reasoning effort to 'low', 'medium', 'high' if gemini, how to allot thinking budget tokens. If this feature is not available, please bring this feature up. It would be useful