OPENAI TESTS NEW “THINKING EFFORT” FEATURE FOR CHATGPT, PROMISING MORE CONTROL OVER ÁI REASONING

OpenAl is experimenting with a new feature for ChatGPT called “Thinking effort”, which allows users to adjust how much computational depth the Al applies before delivering a response.

- 

The option, now in limited testing, could give users unprecedented control over the trade-off between speed and accuracy—reshaping how people interact with generative Al in daily tasks.

WHAT “THINKING EFFORT” MEANS

Traditionally, ChatGPT responses are generated in a single pass, balancing performance and computational efficiency based on preset defaults. The new “Thinking effort” feature introduces a slider or toggle-based control that lets users request the model to apply more intensive reasoning when tackling complex questions.

For example, a user could ask for a quick answer to a straightforward question like “What's the capital of Argentina?” and keep the effort low for an instant reply. But when faced with a thorny prompt such as “Compare the fiscal policies of emerging economies in Asia and their long-term impact on debt sustainability,” they could increase the “thinking effort” setting. The model would then spend more computational cycles evaluating reasoning paths before generating a response.

Early testers describe the feature as something akin to giving ChatGPT a “pause to think” moment. The result is not only more detailed answers but also, in some cases, greater factual accuracy and logical coherence.

WHY OPENAI IS ADDING THIS FEATURE