OpenAI expands ChatGPT free access with GPT-5.6 Luna and unlimited text chats

OpenAI is reshaping the free ChatGPT experience around a new default model and a planned removal of text-chat limits. The company announced on August 6 that GPT-5.6 Luna would become the standard model for Free and Go accounts during the week, replacing GPT-5.5 for those users. A second change is scheduled for the following week: unlimited text conversations, governed by safeguards against abuse, together with a Think control for questions that need more reasoning. The timing matters because the model switch is beginning before the unlimited-chat policy and should not be treated as one simultaneous launch.

The planned unlimited access applies to ordinary text exchanges, not every capability available inside ChatGPT. OpenAI said separate limits will remain for file uploads, images and other tools, preserving a distinction between the core conversation interface and more resource-intensive functions. The company also attached abuse guardrails to the promise of unlimited text use. That qualification makes the change broader than a larger message allowance but not an unconditional removal of every control. For Free and Go users, the practical transition therefore has two stages: Luna becomes the default during the week of the announcement, then the text and Think changes are expected the next week.

GPT-5.6 Luna is positioned as the engine for that expanded free experience. OpenAI reported from an internal evaluation that factual errors were 62% less common with Luna than with GPT-5.5 Instant. That figure is a company measurement rather than an independent benchmark, but it explains why access and model choice are being changed together. If the rollout proceeds as described, more users will encounter Luna not by selecting it from a model list, but simply by starting a normal ChatGPT conversation. The default placement gives the model a central role in the product even before text limits are lifted.

Paid users received a different part of the update immediately on August 6. Plus and Pro accounts gained an updated GPT-5.6 Sol chat model aimed at everyday requests such as questions, research, planning, writing and decisions. OpenAI characterized its answers as more compact and robust, and reported that factual errors in its internal evaluation were 68% less common than with GPT-5.5 Instant. The company also added a thinking slider for Plus and Pro, allowing users to adjust how much reasoning the model applies according to the complexity and number of steps in a request.

The controls reveal a tiered approach to reasoning rather than a single identical interface for everyone. Free and Go users are slated to receive a Think button, a direct choice to invoke more reasoning for a difficult question. Plus and Pro users instead have a slider that offers finer adjustment. Both mechanisms make computational effort visible as a product decision, but they differ in granularity. That distinction may help OpenAI offer a simple default to a broad audience while keeping more detailed control within paid plans, even as basic text conversation becomes less constrained for free accounts.

OpenAI also drew a boundary around which Sol deployments changed. The August 6 update applies to the GPT-5.6 Sol version used for chat by Plus and Pro subscribers. The Sol versions used by Work and Codex were not changed as part of this release. That separation is important because a shared model name does not mean every product surface received the same update. Developers and workplace users should not infer a new underlying model from changes visible in consumer ChatGPT, while paid ChatGPT users can expect the refreshed chat behavior and the new reasoning control now.

The rollout leaves its most consequential promise just ahead rather than already completed. Luna's default placement is due during the current week, whereas unlimited text chats and the Think button depend on the following week's release and continued enforcement of abuse protections. The remaining questions concern how smoothly OpenAI can support much heavier free usage and how often guardrails will shape what users experience as unlimited. For now, the announcement marks a clear shift in access: the free product is moving toward unrestricted text conversation, but tool limits, staged timing and model-specific boundaries still define the offer.