Now Openai reveals more of the thought process of the O3-mini model

In particular, in response to opponents Chinese AI Company DeepSeekOpenai changes the newest AI model, O3-miniStep-by-step follows the “Thinking” process.

Thursday, Openai declared This free and ChatGPT, the company’s AI-powered chatBot platform will see a “chain of thinking” showing the model “grounding” steps and answers to questions. According to Openai, the O3-mini configuration of the O3-mini “high reasoning” configuration will see this updated reading of Premium Chatrpt plans.

“We present the updated chain for O3, designed to make it easier for people to understand how the model thinks.” “With this update, you will be able to follow the justification of the model that gives you more clarity and confidence in their responses.”

Thinking like O3-mini as a thorough fact, check themselves before giving the results Avoid some traps normally walking models. Trade, reasoning models are a little longer to come to solutions – usually up to a few minutes.

Deepseek’s R1 modelThe “justification” model along O3-mini lines shows that many AI researchers have a very thoughtful approach. In addition to facilitating the model of the model, substantial steps provide a better user experience in certain situations, helps the model will be on the right or incorrect.

Openai, O3-mini and predecessors, O1 and O1-mini, in part, because of the reasons for competition, he chose not to show full thinking steps for O1 and O1-mini. Instead, users only saw summaries of reasoning steps – Summaries that are sometimes wrong.

When we informed people before the release of the O1-Preview, the COT’s live broadcast is generally the moment of “AHA”, it clarified that it would be a great thing. These are not raw balls, but a big step is approaching and I’m glad we can share this experience with the world. https://t.co/72zpprhmfk

– Noam Brown (@polynoamial) 6 February 2025

Openai does not show full thinking steps, but the company said, “The company” “found a balance,” he can “think freely” and then organize a more detailed summary.

“To improve the clarity and safety, we’ve reviewed the raw word chain, which eliminates any dangerous content, and then simplifies any complex ideas, this post-processed step is the mother who is not English Allows you to get a thinking chain, create a more favorable and kind experience. “

One Reddit last weekKevin Weil, Openai’s Chief Officer, showed that the change came.

“Today we are working on a gang more than we show – (show the model thinking process) there will be much more soon,” he said. “We also know that they want to have a competitive distillation, which shows all the chains of everything in TBD, but they want people (at least power users), so we will find the right way to balance it.”