OpenAI Is Testing Advanced ChatGPT User Controls
Recent updates to OpenAI's platform suggest the company is experimenting with powerful new tools designed to give users more granular control over their interactions with ChatGPT. Currently being tested with a select group of Pro and Enterprise users, these features focus on adjusting the model's computational effort and managing conversation histories more flexibly.
Deeper Control with a New Effort Selector
One of the most significant changes is a redesigned selector for what OpenAI internally calls "juice" or "thinking effort." This control appears to be moving to a more prominent position in the user interface, allowing for quick adjustments. The update introduces more detailed levels of control over how much computational power is dedicated to a given task. According to code inspections, these levels include "Light thinking," "Standard thinking," "Extended thinking," and a "Max" setting reportedly capped at 200 juice. While there's no official timeline for a wider release, this experiment points toward a future where users can balance response speed against analytical depth.
The new ChatGPT web app version has an updated (hidden) thinking effort picker - Max thinking (200), Extended thinking (48), Standard thinking (18), Light thinking (5)
And a few other related experiments, including showing models in the plus menu, showing the selected model in… pic.twitter.com/onWcBq4Cuw
— Tibor Blaho (@btibor91) August 29, 2025
Explore Possibilities with Conversation Branching
In addition to effort control, OpenAI is also piloting a conversation branching system. This functionality allows a user to return to any previous message in a chat and start a new, distinct line of conversation from that point. This is a game-changer for researchers, developers, and power users who need to explore multiple outcomes from a single prompt. It eliminates the friction of starting over when a conversational path leads to a dead end, making it easier to compare different approaches and retrace steps.
"Branch from here" option has been added to the ChatGPT web app in the latest build (hidden/work-in-progress, probably) - allowing branching of a conversation to a new conversation after a response pic.twitter.com/LVM0l2j5eX
— Tibor Blaho (@btibor91) August 27, 2025
A Response to User Demand for More Control
These experimental features, discovered by inspecting new software builds, reflect OpenAI's ongoing effort to make its advanced models both more powerful and more manageable. The introduction of fine-tuned effort controls and conversation branching seems to be a direct response to user feedback calling for greater transparency and usability, especially as models like GPT-5 become more complex. If rolled out broadly, these upgrades could significantly streamline workflows for teams using ChatGPT for research, content creation, and complex problem-solving.
