OpenAI's "28 Days of Continuous Releases" stunt hit Day 3 with its biggest consumer move yet: the GPT-6 model series is now rolling out globally to every ChatGPT user, free and paid. Plus and Team subscribers get GPT-6 Sol, while free and Go users will see GPT-5.6 gradually replaced by GPT-6 Luna. Alongside the rollout, OpenAI says Codex and Work weekly active users have passed 40 million, with usage quotas reset across the board. After weeks of staged announcements, this is the moment GPT-6 stops being a press release and starts being the default experience for hundreds of millions of people.
The headline feature is Intelligent UI, and it fundamentally changes what a chat response looks like. Instead of plain text, GPT-6 streams interactive cards in real time — checklists, native buttons, dynamic charts, bill splitters, adjustable budgeting tools — rendered on the fly according to the task at hand. Ask for a trip plan and you get an interactive itinerary, not a wall of prose. Under the hood, a technique OpenAI calls Interleaved Thinking cuts first-token latency by 44%, with the model reasoning in real time as it generates rather than thinking first and answering second. The chatbox, the defining interface metaphor of the last three years, is being dismantled in favor of something closer to a dynamic app renderer.
Not everyone is cheering. OpenAI's own system card acknowledges statistically significant regressions on self-harm and extremism evaluations compared with GPT-5.6 — an eyebrow-raising admission to ship alongside a mass rollout to free-tier users, and one that deserves more scrutiny than it has received. The developer crowd, meanwhile, is prickly about the formatting itself: one widely echoed reaction called the heavily structured output "condescending," saying it "treats you like a child," and several observers pointed at OpenAI's newly announced in-chat advertising format as the likely motivation. All that new visual real estate has to be monetized somehow, and interactive cards are a much friendlier container for sponsored content than a monospace text block.
If you build on OpenAI's platform, the practical takeaway is that the chat surface is moving under your feet. Interactive cards mean new UX patterns to learn and new possibilities for ChatGPT-native distribution — but they also mean less control over how your product's output renders inside someone else's interface, and a new dependency on OpenAI's rendering decisions. Before routing anything sensitive through the new models, read the system card's evaluation section yourself rather than trusting the launch narrative. A prettier interface is not the same as a safer model, and the gap between those two things is exactly where production incidents come from.
