OpenAI is testing a new feature called “Thinking effort” for ChatGPT. The feature is likely to allow users control over how much reasoning power the AI applies when generating responses. The feature was first spotted by X user Tibor Blaho in the latest version of the ChatGPT web app.
A hidden menu reveals four levels of thinking effort: Light thinking (5), Standard thinking (18), Extended thinking (48), and Max thinking (200). Obviously, OpenAI has not revealed any details about what these numbers represent, but they likely indicate the number of internal reasoning steps the model takes before producing an answer.
The screenshot shared by Blaho shows the option situated inside the model picker, where users can manually choose how deeply ChatGPT should “think” before responding. The company also appears to be experimenting with related tweaks like showing models in the Plus subscription menu, highlighting the currently selected model in the composer, and introducing a fully collapsed tool menu.
However, this is not the only new feature the company is working on. Recently, another experimental feature called “Branch from here” was spotted in ChatGPT. This tool allows users to split a conversation into a new thread while keeping the original intact.
So, with all these new features in the testing or development phase, one thing is clear: in the coming time, we are going to witness a spree of new feature launches on ChatGPT. However, when these features will arrive, that is something only time will tell. But until that time, if you want to stay updated with all the latest AI updates, then join us on WhatsApp.
