OpenAI is changing the way its o3-mini model communicates its step-by-step “thought process.
OpenAI has announced that free and paid users of ChatGPT will see a different “thought process” with more reasoning. So the o3-mini model shows how it arrived at the answers to the questions.
Detailed summary
“We are introducing an updated thought chain for o3-mini that is designed for people to more easily understand how the model thinks,” an OpenAI spokesperson told TechCrunch. Reasoning models like o3-mini check themselves thoroughly before giving results, thus avoiding pitfalls. One downside to that trade-off is that reasoning models usually take seconds to minutes longer to generate an answer.
Source: OpenAI
OpenAI does not show all of o3-mini’s reasoning steps due to concurrency reasons. Instead, we are shown a detailed summary of o3-mini’s “thoughts. DeepSeek, on the other hand, does show the full thought process of its R1 model. This makes the model easier to study and provides a better user experience. OpenAI’s spokesperson states, “This update allows you to follow the model’s reasoning, giving you more clarity and confidence in the responses.”
The updated version is available to free and paying users of o3-mini. Premium ChatGPT subscribers using o3-mini in the “high reasoning” configuration will also see this more detailed process.