OpenAI is scrapping the daily message caps that have long limited how much free users can chat with ChatGPT, a change that arrives just as the assistant crosses a headline milestone. The company said text-based conversations will become unlimited across its tiers, framing the move as a response to reaching more than a billion weekly users. The shift, announced on August 6, 2026, retires one of the most familiar friction points of the free product and reframes generous access as a growth strategy rather than a cost to be rationed.
What changes for free ChatGPT users
For people on the no-cost plan, the shift removes one of the most visible frustrations of the service. Until now, free accounts hit a ceiling after a set number of messages in a rolling window and had to wait before continuing, a limit that pushed heavy users toward paid subscriptions. Reporting described the previous free cap as roughly ten messages every five hours, a wall that regular users learned to plan around. Making text chats unlimited erases that barrier for ordinary back-and-forth conversation, which is the way most people use the tool day to day, and the company said the change requires no credit card to unlock. In practice, that means a student working through a long study session or a job seeker drafting and redrafting a cover letter no longer runs into a wall mid-task and gets pushed into a waiting period, the kind of interruption that most often nudged casual users to abandon a conversation or reach for a competing assistant.
The company rolled the change out in August 2026, with the update beginning the week of August 10 and reaching free and lower-cost accounts first before extending across the user base. The announcement, detailed by The AI Insider, arrived alongside a set of new default models. Coverage from TechCrunch noted that the unlimited text access was tied directly to the service passing the one-billion-user mark.
The billion-user milestone driving the shift
The timing is not incidental. Crossing a billion weekly users gives OpenAI a scale that reshapes the economics of the free tier, since a service that large can afford to treat generous access as a growth engine rather than a cost to ration. Removing limits also serves a competitive purpose, keeping casual users inside ChatGPT rather than sending them to look at rival assistants once they hit a wall. At that scale, retention and habit matter as much as any single feature, and every conversation that stays inside the product deepens the moat against competitors chasing the same audience. A larger free base also feeds the paid funnel, since people who lean on the assistant daily are the ones most likely to eventually pay for heavier features.
The GPT-5.6 Luna and Sol models behind it
The change is paired with a model refresh. A new model described as GPT-5.6 Luna becomes the default for free and lower-cost accounts, replacing the previous GPT-5.5 generation, while paying subscribers on the Plus and Pro plans gain access to an upgraded model described as GPT-5.6 Sol tuned for quick tasks such as questions, research, planning and writing. Reporting noted that Sol now handles both fast replies and longer reasoning through a single model, rather than forcing subscribers to switch between separate instant and thinking modes. Lowering the compute cost of each answer through more efficient models is part of what makes unlimited text chat feasible for such a large audience, since a cheaper response per message is what turns a costly promise into a sustainable one. Naming the models after celestial bodies, Luna for the free default and Sol for the paid tier, also gives the company a cleaner way to signal capability tiers than the string of version numbers that came before, a small branding shift layered on top of the technical one.
Which limits OpenAI is keeping
The word unlimited applies to text, not to everything. The company said separate caps will stay in place for files, images, voice and image generation, the features that consume far more computing power than a plain text exchange. That distinction keeps the most resource-intensive uses metered while opening up the lightweight conversation that dominates everyday activity, a balance that lets the company advertise unlimited access without absorbing runaway costs. Reporting also flagged an unchanged default that often goes unnoticed: conversations still help train the next model unless a person turns that setting off, so the trade for free, uncapped access remains the data those chats generate.
The new Think button and error-rate claims
Alongside the limit change, OpenAI introduced a Think button that routes harder questions to a slower, more deliberate reasoning mode when a quick answer will not do, giving even free users a way to summon extra effort on a per-message basis. The company also cited internal testing to argue the new models are more accurate, saying factual errors were 62 percent less common in the Luna model and 68 percent less common in the Sol model than in the earlier GPT-5.5 version. Those figures come from OpenAI’s own evaluations rather than independent audits, so their real-world reliability will become clearer as a billion users put the models to work. The rollout was also paired with new safeguards aimed at teen accounts, a sign the company is weighing the risks of far wider, uncapped access alongside its benefits. Taken together, the package amounts to more than a single feature tweak: a cheaper default model, an uncapped free tier, a reasoning shortcut and fresh guardrails all shipped at once, aimed at turning a billion weekly users into a durable daily habit rather than an occasional stop.
This article was produced with the assistance of AI and reviewed by Morning Overview editors prior to publication.
More from Morning Overview
