GPT-6 Intelligent UI is rolling out across ChatGPT, giving the service a way to build interactive answers with charts, buttons, forms, diagrams and task-specific tools. OpenAI began the global release for paid plans on October 7, with Free and Go access scheduled to start October 8.

The rollout introduces four concrete changes:

• ChatGPT can generate interactive interfaces inside answers.
• GPT-6 can respond progressively while it continues working.
• Paid Chat plans use GPT-6 Sol; Free and Go use Luna.
• ChatGPT Work and Codex models remain unchanged.

More on This Story

• OpenAI Launches Dots Always-On AI Agents for ChatGPT
• Google Unveils Gemini 4 Argon With 1M-Token Output Limit

GPT-6 Intelligent UI Builds Interfaces Inside Answers

OpenAI’s announcement describes Intelligent UI as a response system that combines text, visuals and interactive controls according to the question. The model can still return plain text when that is the clearest format, rather than forcing every reply into a graphical layout.

The examples range from side-by-side comparisons and explorable diagrams to calculators, bill splitters and simple games. A planning request can produce a timeline or map, while a learning prompt can create controls that let a user change an input and inspect the result.

This is different from attaching a static image to a conventional answer. OpenAI says GPT-6 decides which components to use and how to arrange them. The interface is generated as part of the response and can accept interaction without sending the user to a separate application.

The feature relies on a library of native components and a compiler that processes the interface as the model produces it. That approach gives generated experiences a common visual foundation while allowing the model to vary their content and structure for each request.

The compiler also lets elements appear progressively instead of waiting for an entire interface to finish. OpenAI trained the model to make choices about layout, completeness and when interactivity is useful, although the company acknowledges that its design judgment still needs improvement.

Sol and Luna Divide the ChatGPT Rollout

GPT-6 Intelligent UI started rolling out globally in the Chat tab to Plus, Pro, Business and Enterprise customers on October 7. Enterprise access depends on workspace administrator settings. Free and Go accounts are scheduled to join the rollout from October 8.

The underlying model depends on the plan. Plus, Pro, Business and Enterprise use GPT-6 Sol for the standard Chat experience, while Free and Go use the more efficient GPT-6 Luna. Intelligent UI is available across reasoning levels from Instant through Extra High.

There are important boundaries. The Pro reasoning option continues to use GPT-6 Astra and does not support Intelligent UI. The October update also does not replace the models used in ChatGPT Work or Codex, which remain on versions released for those products separately.

The Verge reported that the rollout brings the Sol and Luna versions into ChatGPT after OpenAI introduced the first GPT-6 models in September. That makes the October release a broad product deployment and interface change, rather than the first appearance of the GPT-6 family.

GPT-6 Starts Answering Before Work Is Finished

The new Chat experience can begin presenting useful material while GPT-6 continues reasoning or using tools. Later portions of the answer can add findings without requiring another prompt. The goal is to reduce the time before a user can begin reading while preserving a coherent final response.

OpenAI says GPT-6 Instant began answers to search-dependent questions 44% sooner on average than GPT-5.6 Instant in an internal evaluation. It also says GPT-6 Extra High started responding as quickly as GPT-5.6 Medium while earning a better overall score than GPT-5.6 Extra High on everyday agentic tasks.

Those figures come from OpenAI’s own tests. The company did not publish a complete independent benchmark for Intelligent UI, and response speed will depend on the prompt, tools, network conditions and generated components. Progressive output also creates a product challenge: early information may arrive before every source or calculation has been checked.

The interface layer could make ChatGPT more useful for tasks that previously required a spreadsheet, calculator or small web application. Its value will depend on whether generated controls behave correctly, remain accessible on different devices and preserve context when users modify inputs.

OpenAI Classifies GPT-6 as High Capability

OpenAI’s October system card classifies GPT-6 Sol and Luna as High capability in cybersecurity and biological or chemical domains, but below Critical. Neither model reached the company’s High threshold for AI self-improvement.

OpenAI reports stronger resistance to multi-turn jailbreak attempts, along with reductions in deception and guardrail circumvention compared with GPT-5.6. It also recorded statistically significant regressions on some standard safety evaluations, including self-harm for Sol and several content categories for Luna.

The company says manual review found those failures generally low severity and that adversarial testing did not reveal high-severity risks under its criteria. Those are company assessments, and the rollout to a very large consumer audience will provide a broader test of both the model safeguards and the reliability of generated interfaces.

OpenAI has not announced separate pricing for Intelligent UI because it is distributed through existing ChatGPT plans. The near-term questions are how consistently the model chooses a useful format, whether interactive tools calculate correctly and how quickly the feature reaches eligible accounts across web and mobile.