OpenAI now reveals more of its o3-mini model’s thought process
In response to pressure from rivals, including Chinese AI company Deepseek, OpenAI changes the way in which the latest AI model, O3-Mini, communicates its step-by-step ‘thought’ process.
On Thursday, OpenAi announced That free and paid users of Chatgpt, the AI -driven chatbot platform of the company, will see a updated “thought” that shows more of the “reasoning” steps of the model and how it came to answers to questions. Subscribers of Premium ChatGPT plans that use O3-Mini in the “High reasoning” configuration will also see this updated reading, according to OpenAI.
“We introduce an updated [chain of thought] For O3-Mini who is designed to make it easier for people to understand what the model thinks, “a OpenAi spokesperson told Techcruch via e-mail. “With this update you can follow the reasoning of the model, giving you more clarity and trust in his answers.”
Reasoning models such as O3-Mini check themselves thoroughly before they give results, which helps them to avoid some of the pitfalls that normally stumble on models. The assessment is that reasoning models take a little longer to come to solutions to minors to minutes longer.
The R1 model of Deepseek, a “reasoning model” in the style of O3-Mini, reveals its full thinking process, which, according to many AI researchers, is the preferred approach. In addition to making the model easier, the reasoning steps yield a better user experience in certain situations, so that they can indicate when the model is on the right – or wrong.
OpenAi had chosen not to show the full reasoning steps for O3-Mini and its predecessors, O1 and O1-Mini, partly for competing reasons. Instead, users only saw summaries of the reasoning steps – summaries that were sometimes incorrect.
OpenAi still does not show the full reasoning steps of O3-Mini, but the company said it has “found a balance”: O3-Mini can “free thinking” and then organize his “thoughts” in more detailed summaries.
“To improve clarity and safety, we have added an extra step after processing whereby the model assesses the unprocessed line of thought, any unsafe content is removed and then simplifies all complex ideas,” the OpenAi spokesperson continued. “Moreover, this post-processing step does enable non-English users to receive the idea in their mother tongue, creating a more accessible and friendly experience.”
In a Reddit Ama last week, Kevin Weil, the main product officer of OpenAi, shed through that the change was coming.
“We are working on to show some more than we show today – [showing the model thought process] Will be very, very quickly, “he said. “TBD about everyone – showing all line of thought leads to competitive distillation, but we also know that people (at least powerful users) want it, so we will find the right way to balance it.”
