OpenAI is working on a new feature for ChatGPT, called Thinking effort picker.
With this feature, you can choose how hard ChatGPT thinks. However, just because a model thinks harder doesn't necessarily mean it will give a better answer.
In which cases can the Thinking effort picker be useful?
The new ChatGPT feature can be useful when dealing with a complex topic, such as econometrics, bond valuation, healthcare , etc.
See also: ChatGPT-5: Bypassing Security in a Few Words
As spotted on Platform X (formerly Twitter), OpenAI is testing an effort picker, so you can choose the “thinking intensity” of the model. There are currently four levels. The lowest level is called light and has an internal attribute of 5, while standard is at 18, extended at 48, and max at 200.
This number represents an internal “juice budget.” More “juice” means the model takes more thinking steps and typically provides deeper answers, but responds more slowly.
The maximum thought (200) is limited — meaning it will only be available to those with the $200 Pro subscription.
While the details remain unclear, BleepingComputer claims that OpenAI is likely looking to give users more control, as long as they understand how a model works in relation to the types of questions you ask.
See also: Attention! ChatGPT chats appear via Google search
For example, you can choose the light for quick questions, the standard for everyday use, the extended for moderately complex issues, and the max when you need the most careful and in-depth analysis.
OpenAI ChatGPT
The Thinking effort picker represents an attempt by OpenAI to transfer some of the control of computational effort directly to the user, something that has so far only been done “behind the scenes” of large language models. The idea that one can decide how intensively ChatGPT will “think” is novel, as it introduces a dimension of parameterization that is more reminiscent of choosing the level of detail in analytics software or even scaling power in computing clusters.
On a practical level, this feature can help professionals in fields where depth of analysis is critical. A financial analyst, for example, might prefer the extended or max for risk assessment scenarios, while a student looking for a short article summary might be satisfied with light. In other words, the new option is not just about speed, but also about the degree of mental investment the model puts into the answer.
There is, however, an interesting dimension to the user experience: the ability to control may create the illusion that “more thought” always equals “better quality.” This can lead to false expectations, especially when the subject of the question does not require complex reasoning. In such cases, overusing max may simply produce more extensive but not essentially better answers.
See also: ChatGPT lets you choose different personalities
Furthermore, the existence of the “juice budget” introduces a new cost-performance relationship. On the one hand, it allows demanding users to have access to more in-depth analysis; on the other hand, the exclusion of the highest level in the expensive Pro package ($200) shows that OpenAI wants to experiment with a premium “high cognitive effort” experience. This is expected to fuel a discussion about whether knowledge and analytical power should be priced at different levels.
Ultimately, the Thinking effort picker is not just a technical setting. It is a tool that invites the user to consider their needs: do they want speed and brevity or depth and analytical rigor? The answer to this dilemma may also determine how ChatGPT will be integrated into everyday life, from the office to education.
Source: www.bleepingcomputer.com
