Chatbots have long forced a quiet trade-off between speed and depth, answering trivia in an instant but sometimes rushing through problems that deserve more deliberation. OpenAI’s newest update to ChatGPT tries to hand that choice to the person at the keyboard, letting them decide in the moment how much thinking the assistant should do. The change is small in appearance, a single control tucked into the interface, but it reflects a larger rethinking of how much say a person should have over the machinery doing the reasoning.
The reasoning-effort slider
The centerpiece of the August 2026 update is a control that adjusts how much internal reasoning the model performs before it replies. The slider appears across the web, mobile, and desktop versions of ChatGPT, and it went out to paying Plus and Pro subscribers alongside the refreshed model.
Functionally, the dial governs how many tokens the system spends on planning and working through a problem before producing an answer. According to coverage from 9to5Mac, a low setting suits quick, everyday questions where a fast reply is fine, while a higher setting is meant for research, coding, writing, and other tasks where careful, step-by-step reasoning pays off.
GPT-5.6 Sol and the end of mode-switching
The slider arrived with an upgraded model that OpenAI calls GPT-5.6 Sol. Its notable change is structural: a single model now handles both snap responses and extended reasoning, replacing the earlier arrangement in which separate systems powered distinct “Instant” and “Thinking” modes. As The Decoder reports, folding those behaviors into one model is what makes a continuous dial possible in the first place.
The practical effect is that adjusting effort feels less like jumping between two different personalities and more like stretching or shortening a single line of thought. The same underlying model simply spends more or less time reasoning, which the company argues yields more consistent tone and behavior across a range of tasks.
Why exposing the effort setting matters
For most of the chatbot era, the amount of hidden reasoning behind an answer was fixed by the provider and invisible to the person asking. Surfacing that control acknowledges a real tension in how these tools are used: extra deliberation improves hard answers but costs time and computing power, and it is wasteful on simple requests. Putting the knob in the open lets someone match the effort to the job rather than accept a one-size-fits-all default.
The design also has an economic logic. Reasoning tokens are expensive to generate, so allowing lighter settings for casual questions can trim the cost of running the service, while reserving heavy computation for the moments that genuinely benefit from it. It also gives heavier users a lever they had long requested, since developers and researchers often know in advance whether a problem warrants deep reasoning and prefer not to guess at what the model will do behind the scenes.
Where the feature fits in a broader rollout
The dial was not entirely new to the ecosystem. A version of it had already appeared in ChatGPT Work, the business-tier offering, before the broader consumer release brought it to individual subscribers. That staged approach mirrors how OpenAI often tests interface changes with enterprise customers before pushing them out widely, gathering feedback from heavier users first.
What the effort setting means in practice
For everyday use, the practical effect is a trade between speed and thoroughness that the person can judge for each task. A quick factual question or a casual rewrite rarely benefits from extended deliberation, so a low setting returns an answer almost immediately, while a thorny coding bug or a piece of analysis can be handed a higher setting and given room to work through intermediate steps.
The design also nudges people to think about how much reasoning a task actually deserves, a judgment that used to be made invisibly by the software. Over time, matching effort to difficulty could reduce wasted computation on trivial requests while improving results on the hard ones, assuming the control is used deliberately rather than left at a single default.
The trend toward tunable thinking
OpenAI is not alone in experimenting with adjustable effort. Rival AI developers have introduced their own buttons and toggles that let people trade speed for depth, reflecting an industry-wide realization that a fixed level of reasoning rarely satisfies everyone. As these models grow more capable and more costly to run, giving people a direct lever over how hard the system works may become a standard feature rather than a novelty, reshaping the simple chat box into something closer to an instrument with a volume control for thought. For now the dial is a modest addition, but it marks a shift in philosophy, treating the amount of thought a model applies not as a fixed property of the software but as a setting worth handing to the person who knows the task best.
This article was produced with the assistance of AI and reviewed by Morning Overview editors prior to publication.
More from Morning Overview
- Card skimmers hidden on gas pumps and ATMs are draining accounts, and here’s the tell
- The FBI says hackers are hijacking outdated home routers, and it named the models to check
- Older Teslas are wearing out in ways early owners never saw coming
- A common childhood virus is now tied to multiple sclerosis years later
