OpenAI is testing "Thinking effort" for ChatGPT

OpenAI is testing a feature called the Thinking effort picker for ChatGPT, allowing users to choose how hard the model thinks. There are four levels: light, standard, extended, and max, with internal juice budgets of 5, 18, 48 and 200 respectively. More juice means the model takes more steps and usually provides a deeper answer, but it also responds more slowly. The max thinking level is gated and available only to those with a $200 subscription (Pro plan). The information was reported on 31 August 2025, with OpenAI’s test noted as a way to give users more control over the model’s thinking intensity.