docs: use 002 model for better scalability in text generation

docs: use 002 model for better scalability in text generation #6103