GPT-5 Thinking can now think in different modes or levels.
OpenAI introduces a new feature that allows users to determine how deeply the GPT-5-Thinking mode reasons. The option is intended to give users more control over the speed and depth of responses.
Choice in Thinking Levels
According to BleepingComputer, OpenAI initially experimented with a slider, but ultimately opted for a simple list of different thinking times, internally referred to as “juice” values. The default setting gives GPT-5-Thinking a juice level of 18, providing users with a balance between speed and intelligence. The ‘extended’ mode increases this level to 64, for more thoughtful and complex responses.
Pro users get two additional modes: Light (juice level 5) for quick but superficial responses, and Heavy (juice level 200) for maximum thinking power. The latter setting is intended for very difficult questions. In our test, GPT-5 proves to be a structured and business-like thinker. Unlike its emotional predecessor GPT-4, GPT-5’s responses are more matter-of-fact and less excessively positive.
More Control over Performance and Speed
With this new feature, OpenAI aims to give users more control over how the model should reason: quickly or intelligently. The box is visible for Plus and Business accounts, and for Pro accounts also with the additional Light and Heavy modes.