OpenAI is removing limits on text-based chats for all ChatGPT users, meaning you can now type as many messages as you want without hitting a cap. The change was announced on August 6, 2026, alongside a new default model called GPT-5.6 Luna for free and Go users (replacing GPT-5.5) and an upgraded model called GPT-5.6 Sol for paying Plus and Pro subscribers (TechCrunch, OpenAI).

The announcement comes just after ChatGPT passed 1 billion weekly active users (TechCrunch). The removal of text-chat caps applies to everyone, but limits on other features — uploading files, generating images, using voice, and creating images — still remain (TechCrunch).

For free and Go users, the new default model is GPT-5.6 Luna. These users also get a new "Think" button that lets the AI spend more time working through hard questions before answering. Think of it like asking someone to pause and reason through a problem rather than blurting out the first answer that comes to mind. For Plus and Pro subscribers, OpenAI is shipping an upgraded GPT-5.6 Sol model designed for quick tasks like answering questions, doing web research, giving advice, planning, writing, and decision-making. This version of Sol is separate from the one used in Codex and Work environments, which stays the same (TechCrunch, OpenAI).

Paying users also get a slider that controls how hard the AI thinks before responding. The free-tier Think button and the paid-tier slider do the same basic thing — they let you decide whether you want a faster answer or a more carefully reasoned one — but the slider offers finer control (TechCrunch, OpenAI).

OpenAI's own testing reports fewer factual errors compared to the previous model: 62% fewer for GPT-5.6 Luna and 68% fewer for GPT-5.6 Sol (TechCrunch). These are OpenAI's own results, not independent tests, and the company has not shared the specific questions or methods used to measure this.

The rollout is staggered. The upgraded GPT-5.6 Sol is available to Plus and Pro users right away. Changes for free and Go users, including the Luna model, arrive during that week. Unlimited text chats and the Think button for free and Go users come the week after (TechCrunch).

Two other items appeared on OpenAI's newsroom the same day. The company published a post titled "Improving GPT-5.6 Sol in ChatGPT—and expanding access to GPT-5.6 Luna for free users" on August 6, and a partnership post titled "OpenAI and APA partner to advance responsible AI" with the same date. An engineering post on "Continuous voice interaction with GPT Live" was dated August 3 (OpenAI Newsroom.

Up to now, OpenAI has used message limits as the main way to separate free and paid users. As of a May 2024 support page, ChatGPT Plus users had message limits up to five times greater than free users, with Team and Enterprise tiers higher still (OpenAI). The free tier's capabilities, default model, and limits have always changed over time (OpenAI Help).

Removing text-chat caps changes that approach. The free tier now has limits only on features like files, images, voice, and image generation, while text conversations are effectively unlimited. This shifts the reason to pay from "more messages" to better models (Sol versus Luna), more control over reasoning (the slider versus the Think button), and higher limits on those other features.

There is a detail here that matters for developers and businesses. The upgraded GPT-5.6 Sol going to Plus and Pro users is a version built specifically for ChatGPT, tuned for everyday tasks like advice and planning. The version of Sol used in Codex and Work — OpenAI's tools for coding and enterprise use — is deliberately left alone. Teams using Sol through API or Work interfaces will not see any changes, which should help avoid disruptions in existing systems even as the consumer product shifts.

The error-reduction numbers, while eye-catching, follow a pattern we have seen before in AI releases. Companies report improvements from their own testing that are hard to judge without seeing the test questions or methods. A 62% drop in factual errors sounds large, and it may be, but the baseline model and the definition of what counts as an error are not publicly spelled out. Anyone deciding whether to switch to Luna or the new Sol should treat these figures as a general guide rather than proof until outside testing confirms them.

For the billion people now using ChatGPT each week, the practical change is simple. Text conversations no longer stop at a limit. A Think button lets you ask for deeper reasoning on tough questions. The default model is Luna, which OpenAI says is more accurate than before. Whether that accuracy holds up when a billion people are asking it everything under the sun is a question the rollout itself will answer.