On August 6, 2026, OpenAI announced that it will make GPT-5.6 Luna the default model for ChatGPT's Free and Go plans, and will lift limits on text conversations. The switch to Luna begins this week, while the removal of usage limits and the new "Think" button will roll out next week. Previously, usage of GPT-5.5 Instant on the free tier was capped within 5-hour windows. With this change, ChatGPT moves closer to being a service users can rely on for continuous text-based conversations without interruption. Meanwhile, limits on images, files, and various tools remain unchanged, so misunderstanding the scope of "unlimited" could lead to a gap between expectations and reality.
What Becomes "Unlimited" Is Text Conversation
The scope of the unlimited offering is text conversations using GPT-5.6 Luna. OpenAI states that it will switch the default model for Free and Go plans to Luna sometime this week, and will remove text rate limits starting the following week. This isn't a matter of users selecting Luna from a model list—in standard ChatGPT, Luna will automatically handle default responses for the free tier.
However, limits remain in place for file uploads, images, and other tools. Guardrails to prevent misuse will also continue to apply, so this isn't a guarantee that covers automated, high-volume sending. OpenAI has not disclosed service-level commitments regarding response speed or throughput during peak congestion; what's promised here is only that text conversations won't be cut off due to usage caps.
The "Think" button gives Luna more time to reason through difficult questions. It isn't a feature that switches free users to GPT-5.6 Sol—rather, it's an operation that extends Luna's own reasoning time. Sol is not included in the Free or Go plans, and Luna itself cannot be directly selected from the standard chat model list. This distinction means that while continued use of the free tier is expanding, access to the top-tier model has not been opened up.
From 5-Hour Windows to Always-On Luna
OpenAI's Help Center has explained that GPT-5.5 Instant usage on the free tier is capped within 5-hour windows. The cap isn't a fixed value—it varies based on market and system conditions, anti-abuse measures, and individual usage patterns. After this update, Luna will become the default for Free and Go plans, and at least text conversations will no longer be blocked by this time-window restriction.
This change will matter more to users who want to continue working while maintaining context, rather than those who send just a few short questions. Tasks like polishing a résumé or long-form writing, ongoing Q&A while studying, or fleshing out travel plans can have their flow disrupted if the usage cap is hit midway through. With text-based limits removed, it becomes easier to keep revising within the same thread. According to OpenAI, ChatGPT now has 1 billion weekly users, so this change to the free tier will reach a much broader audience than the addition of a paid feature would.
The boundary with paid tiers has also been redrawn. On Plus and Pro, GPT-5.6 Sol—tuned for everyday conversation—handles everything from instant responses to deep reasoning, with a slider to control how much thinking is applied. On Free and Go, the low-cost Luna is the default, with Think available to press when needed. The gap in whether text conversations can continue uninterrupted has narrowed, leaving the main difference with paid tiers in the capability of the models available and the amount of reasoning applied.
Free Tier Expansion Arrives One Week After an 80% Price Cut
Ahead of this free-tier announcement, on July 30, 2026, OpenAI cut Luna's API pricing by 80%. Pricing dropped from $1 to $0.20 per million input tokens, and from $6 to $1.20 per million output tokens. Luna is described as the fastest and cheapest model in the GPT-5.6 lineup, designed to handle high-frequency processing use cases.
API pricing isn't a figure that directly represents OpenAI's internal cost per ChatGPT conversation. Still, the fact that Luna's price cut and its rollout to the free tier happened within a 7-day span aligns with a strategy of having a low-cost model absorb the bulk of everyday processing. The company has explained that it has made the systems connecting models, inference infrastructure, and tools more efficient, allowing more processing to be completed with the same computing resources.
With this allocation, there's no need to make the expensive Sol the standard for all users. Luna can handle the bulk of everyday conversations, while paying users get access to Sol's advanced reasoning capabilities and adjustable settings. OpenAI has expanded the free tier while still assigning models with different capabilities and costs across its various plans.
How Much Will Free-Tier Response Quality Improve?
According to OpenAI's internal evaluations on detailed fact-based questions in finance, medicine, and law, responses from GPT-5.6 Luna contained at least one factual error about 62% less often than those from GPT-5.5 Instant. The company attributes this to improvements in the ability to use retrieved information in a way that's tailored to the question. This suggests that as usage volume on the free tier increases, the accuracy of the default model is also improving.
That said, the announcement page doesn't disclose the number of evaluation cases or a breakdown of the problem types, and the 62% figure is a relative value from OpenAI's internal testing. It's not a number that guarantees medical or legal answers are correct. The need to verify sources for important decisions remains.
The real test will come after next week's rollout, when it becomes clear whether context retention and response speed hold up during extended conversations on free accounts. For tasks that heavily involve images or files, separate usage caps will kick in first. If Luna remains stable even under peak congestion for text-centered consultation and writing, the free version of ChatGPT will take a step forward from being a trial-limited service to becoming part of users' daily workflow.