Is There a Plan to Enable Function Call Support for the gpt-4-vision-preview Model?

see update here: