**OpenAI** began rolling out GPT-6 with “Intelligent UI” on Wednesday (local time), marking the start of ChatGPT providing its more than 1.2 billion weekly users with answers that go beyond plain text to include charts, buttons, and interactive tools.
Key points
GPT-6, featuring built-in Intelligent UI, began rolling out to paid ChatGPT plans on Wednesday and will expand to the Free and Go plans starting Thursday.
The new model combines text with graphics, input forms, and small tools in its responses, while still providing plain-text answers as before when appropriate to the question.
An early review warns that interactive responses can consume tokens faster and that visual elements may lead to new forms of errors and hallucinations.
GPT-6 launch details
OpenAI announced GPT-6 in a blog post, explaining that the model was trained to combine text, visual elements, and interactive components to suit each question. ChatGPT responses can now include graphics, clickable buttons, input forms, charts, and more.
When a question is best answered briefly, the response will remain plain text as before. Users can also explicitly request tools such as a bill-splitting calculator, a retirement savings calculator, or a simple game.
Paid subscribers got access first. GPT-6 Sol was deployed as the engine powering the Plus, Pro, Business, and Enterprise plans in the Chat tab, while GPT-6 Luna is the model for Free and Go users, with updates scheduled to roll out gradually starting Thursday.
In one demo, ChatGPT broke down a 7-speed bicycle into five systems, including the frame, drivetrain, and brakes. Tapping each system opened a more detailed explanation.
In another example, while planning a Sunday lamb roast party, it presented images of each dish and a cooking timeline, along with a guest counter that automatically recalculated the shopping list based on the number of invitees. OpenAI said GPT-6 uses a library of prebuilt components that work across web and mobile, and that a dedicated compiler renders the interface in stages rather than generating the entire response first and showing it all at once.
Read also: How much does Anthropic’s CEO make? The IPO filing has the answer
The risks of intelligent UI
This approach has precedent.
**Google** introduced generative interfaces last year alongside its Gemini 3 model. An early review of OpenAI’s launch noted that the feature attracted attention when it was unveiled, but did not meaningfully distinguish Google from competing AI labs. The review also warned that interactive responses can use up users’ tokens faster and that visual elements create new opportunities for model hallucinations.
OpenAI GPT-6 launch timeline
OpenAI acknowledged that the model still has room to improve its design judgment. The company described intelligent UI as an intermediate step toward “software that adapts itself to people,” and said GPT-6 Instant starts answering web-search queries 44% faster on average than GPT-5.6 Instant.
This release effectively brings GPT-6’s rollout strategy to its final stage. OpenAI first made its flagship model, GPT-6 Astra, available to select institutional customers on September 3, then added the smaller Sol and Luna models to Work, Codex, and its developer product lineup on September 22. On October 5, it announced plans to begin testing a clearly labeled visual ad format during image generation for Free and Go users in the U.S. later this month.
Next article: OpenAI’s 722-paper AI math collection is fully public… Now it’s up to mathematicians to evaluate it