A company plans to have a large language model create source code from natural-language code comments. Which LLM capability best fits this requirement?

Text generation is the best fit because the task involves creating new source code based on natural language input. The LLM needs to understand the comments and then generate entirely new, functional code. Text summarization condenses existing text, which isn't the goal here. Text completion predicts and fills in missing parts of existing text, but the requirement is to create new code, not just complete a snippet. Text classification categorizes text into predefined labels, which is irrelevant to generating source code.