OpenAI o1

On ToTheMoon, OpenAI o1 marks the point where the channel began treating reasoning models as a class of their own. It comes up in discussions of what a user gets from a system that thinks longer, what that should cost, and how to choose between modes of the same ChatGPT. Over time o1 becomes the benchmark against which DeepSeek, Gemini and paid agents are measured.

In autumn 2024 the hosts examined o1-preview as a change in the way the model works. Episode 024 explained the key difference: the model was trained with full reinforcement learning and searches through options while generating, which is why it costs more and comes with request limits, and why the market is splitting into one model for conversation, another for code and a third for expensive analysis. Episode 025 described o1 as a qualitative leap that had not yet become a product. Episode 036 argued the two-hundred-dollar o1 Pro should be priced against a specialist's labor.

In 2025 o1 became a daily tool and a reference point. In episode 042, against Gemini Flash Thinking and DeepSeek R1, one host said he had seen a completely different level of work on o1 Pro. Episode 043 ran the same prompts through both and found the final answer from o1 in the twenty-dollar tier better than DeepSeek's, even if DeepSeek's visible reasoning read better. In episode 044 a host explained that he runs many tasks for his teams in o1 Pro, and episode 049 named o1 Pro, Deep Research and 4o together as the bar any paid digital employee would have to beat. By episode 063, o1 Pro was already the previous generation next to o3 Pro.