Maybe not even that. It’s ideally a behind-the-scenes performance improvement for LLMs. If it hadn’t been so explicitly marketed as reasoning, people using those LLMs would probably just think it was a slightly better LLM (like the jump from GPT-3 to GPT-4).