OpenAI opens GPT-6 to all ChatGPT users globally

The easiest way to misread this release is to treat it as a simple model upgrade. In fact, OpenAI announced the rollout of GPT-6 to all ChatGPT users worldwide (including free and paid users), replacing GPT-5.6 Sol and GPT-5.6 Luna in ChatGPT. Last month, the first batch of GPT-6 models was already released to paid users, so the essence of this move is a further expansion of coverage, not a from-zero-to-one debut.

The expansion of coverage itself has clear commercial implications: OpenAI says GPT-6 brings the latest generation of intelligence to the more than 1.2 billion people who use ChatGPT each week. It should be noted that this figure comes from OpenAI's own statement, is a source_claim, and lacks independent verification.

Alongside the coverage expansion is a tiered strategy. In ChatGPT, Plus, Pro, Business, and Enterprise users are primarily supported by GPT-6 Sol, while Free and Go users are primarily supported by GPT-6 Luna; both models are optimized for everyday conversation scenarios. Free users get Luna rather than Sol, and the full boundaries of its capabilities are not made clear in the source.

The most closely watched new capability in this upgrade is Intelligent UI, which can quickly generate interactive answers. Unlike the past, when answers relied mainly on text, GPT-6 can now, based on the user's question, autonomously decide how to combine elements such as text, images, charts, buttons, and forms, and directly generate an interactive answer interface.

From a technical implementation standpoint, OpenAI built a native, streaming-generation-capable component library for Intelligent UI, along with a companion interface compiler. When generating an answer, the model can call these components and decide for itself how the elements are combined; the compiler, meanwhile, receives the model's output while progressively rendering the interface, so users do not have to wait for the entire answer to be generated before the interactive interface begins to appear.

To give the model better design judgment, OpenAI expanded its training methods, evaluating the interfaces generated by the model in terms of clarity, usefulness, and completeness. The source also acknowledges that the model still has room for improvement in design judgment and the range of content it can generate—this is a self-limitation from the company, and it is worth keeping in mind when evaluating this capability.

One side effect of reasoning models is waiting: users can have the model tackle harder problems, but they often have to wait for the model to finish thinking before seeing an answer. GPT-6's approach is to be specially trained to interleave thinking and answering, take user waiting time into account, and gradually organize and output answers based on the knowledge and interim conclusions it has at the moment.

The source says GPT-6 has also learned to continuously supplement useful information through multiple interim answers, trying to avoid meaningless filler content while keeping the final answer coherent overall and close to a one-shot complete answer in factual accuracy. This statement is a description of the model's behavior, and its actual effect has not been independently verified in the source.

Two sets of internal evaluation data support this direction. In an internal evaluation of high-value, everyday agent tasks, GPT-6 Extra High's time to start answering was comparable to GPT-5.6 Medium, but its overall score was higher than GPT-5.6 Extra High. For questions requiring web search, GPT-6 Instant started answering on average 44% earlier than GPT-5.6 Instant. Both sets of data come from OpenAI's internal evaluations and are source_claims; the specific methodology, samples, and scoring criteria were not disclosed in the source, and comparisons on third-party benchmarks are likewise unknown.

Availability is being rolled out in phases by user tier, not all at once. GPT-6 with Intelligent UI is rolling out gradually to Plus, Pro, Business, and Enterprise users worldwide through ChatGPT's Chat page starting on the day of release; starting tomorrow, coverage expands further to Free and Go users. Whether Enterprise users can use it depends on the relevant settings configured by their organization's administrator.

The boundaries are equally clear: this update applies only to the Chat experience in ChatGPT, and the models currently used by ChatGPT Work and Codex will not change with this release. Users who access GPT-6 Sol and GPT-6 Luna through Codex and ChatGPT Work will continue to use the previously released versions. Whether GPT-6 will later expand to these two entry points is not addressed in the source.

A few hours before the release, Sam Altman posted a line about looking up at the stars and awe. The source interprets this as an echo at the moment of GPT-6's full rollout, but this is the source's interpretation rather than an official statement and should not be treated as product information.

OpenAI positions Intelligent UI as an important step toward the next stage of ChatGPT: in the future, ChatGPT can dynamically shape an interface that matches the task the user is working on; in the long run, users will only need to describe what they want to accomplish and can then complete the task directly in an interactive experience generated around that goal. OpenAI's phrasing is that for decades people have had to learn how to use software, and the future direction is the opposite—software will begin to adapt to people.

Whether this narrative holds depends on two questions that have not yet been answered. First, the stability and experience of Intelligent UI in real-world use have not been independently verified; second, all performance comparison data comes from OpenAI's internal evaluations, not independent third-party verification. In the absence of external benchmarks, figures like 44% should be understood as vendor claims, not reproducible conclusions.

Therefore, what can be confirmed from this release is the scope, tiering, timeline, and feature boundaries; what cannot be confirmed is the actual magnitude of the capability advantages. Keeping these two types of information separate is the more prudent way to read this full rollout.