OpenAI has begun rolling out GPT-6 across the regular ChatGPT experience, pairing the model upgrade with a new system called Intelligent UI that can turn an answer into an interactive diagram, calculator, form, map or other task-specific interface.
The rollout started October 7 for ChatGPT Plus, Pro, Business and Enterprise users and expands to Free and Go users on October 8. Paid and business plans use GPT-6 Sol in Chat, while Free and Go use GPT-6 Luna. Enterprise access can still depend on workspace settings.
The important change is not simply that ChatGPT can include more images. OpenAI’s product announcement describes a native component system that lets GPT-6 choose and assemble controls inside a response. A comparison can become a side-by-side layout; a mathematical explanation can include adjustable inputs; a dinner plan can add a quantity calculator and checklist. The interface is generated for the request rather than selected from one fixed answer template.
What Intelligent UI can put inside an answer
Intelligent UI supports text, graphics, tappable buttons, forms, charts and interactive experiences. OpenAI’s launch examples include a bicycle diagram with selectable systems, a retirement-savings calculator, a bill splitter, a road-trip map and a cooking plan that changes ingredient quantities with the number of guests.
Those examples reveal a useful dividing line. The feature is best suited to questions where changing one input should update several related outputs, or where the user needs to explore a system rather than read a static explanation. A plain factual question can still receive a conventional text response.
OpenAI built the feature around a library of native, streamable components and a compiler that processes the interface while the model generates it. The component library supplies familiar controls and visual rules; GPT-6 decides what information to place in them and how the parts should fit together. The compiler can begin rendering the interface before the complete response is finished.
That architecture is materially different from asking a chatbot to write an HTML app in a code block. Intelligent UI lives inside the ChatGPT conversation, uses components controlled by the product, and is designed to work across web and mobile. It also gives OpenAI a narrower surface to test than arbitrary generated front-end code.
GPT-6 can answer before all of its reasoning is finished
The same release changes how long-running answers arrive. GPT-6 can begin presenting useful material while it continues to reason or use tools, then extend the response without waiting for another prompt. OpenAI calls this interleaving thinking and answering.
In an internal evaluation cited by the company, GPT-6 Instant began answering web-search questions 44% sooner on average than GPT-5.6 Instant. OpenAI also reported that GPT-6 Extra High could start responding in about the same time as GPT-5.6 Medium while scoring better overall than GPT-5.6 Extra High on its set of everyday agentic tasks. These are vendor-run evaluations, not independent benchmarks, and the company has not published enough detail to translate the numbers into a guaranteed time saving for every conversation.
Progressive delivery creates a new editing challenge for the model: later findings have to fit with what is already visible. OpenAI says GPT-6 was trained to produce a sequence of partial responses that remain cohesive. Users should still treat early material as provisional when a task depends on live research or several tool calls, because the answer can develop as additional sources arrive.
Which ChatGPT plans and modes get the change
The product boundaries are easy to miss because OpenAI now uses related model names across several ChatGPT surfaces.
- Plus, Pro, Business and Enterprise: GPT-6 Sol powers the regular Chat tab as the rollout reaches an account.
- Free and Go: GPT-6 Luna begins rolling out October 8.
- Intelligent UI: Available from Instant through Extra High reasoning where those options are offered.
- Pro reasoning: Continues to use GPT-6 Astra and does not support Intelligent UI.
- ChatGPT Work and Codex: Their underlying models do not change as part of this release.
That last distinction matters for anyone who already selected GPT-6 or GPT-6.1 Sol in Work or Codex. The October Chat release is a separately tuned version for everyday conversation. OpenAI’s release notes also say Intelligent UI is not available when the Pro reasoning option invokes Astra.
Rollout language means two people on the same plan may not see the feature at the same moment. Enterprise administrators can also control whether the model is available to a workspace.
The system card records both gains and regressions
OpenAI’s October GPT-6 system card treats both Sol and Luna as “High” capability in cybersecurity and biological or chemical domains, but below the company’s “Critical” thresholds. Neither model reached its High threshold for AI self-improvement.
The company reported instruction-hierarchy robustness of 99.99% for GPT-6 Sol and 99.79% for GPT-6 Luna in its automated evaluation. It also found fewer cases in which the models tried to work around an action restriction or failed to disclose a blocker. Those tests are useful evidence about the model’s underlying behavior, but they do not eliminate prompt-injection risk in real conversations that use live web content, files or connected tools.
The results were not uniformly better. Relative to their GPT-5.6 counterparts, the October models showed statistically significant regressions in several standard content-safety evaluations. Sol regressed on self-harm; Luna regressed on self-harm, gore and sexual content. In evaluations focused on users under 18, both models regressed on age-restricted content, sexual content and emotional reliance, with Luna also regressing on gore.
OpenAI’s manual review characterized the observed failures as borderline or lower severity and said system-level interventions, including teen protections and crisis resources, sit above the base model. The findings still deserve attention because the regular Chat surface reaches a much broader population than developer tools or paid-only work modes.
What to try first, and what to verify
Users can test Intelligent UI most clearly with prompts that contain variables, stages or competing choices. Ask for a savings calculator with editable monthly contributions, a trip plan with selectable stops and driving times, or a labeled diagram whose parts can be explored one at a time. For a project decision, request a comparison with controls for budget, deadline and priority.
Then check whether the generated interface preserves the facts behind the presentation. A polished calculator can still encode a wrong assumption; an interactive map can still use incomplete or outdated information. For financial, health, legal or safety-sensitive tasks, verify the underlying numbers and sources independently.
The release moves ChatGPT beyond a scrolling transcript toward software that assembles a temporary interface around a question. Whether that improves everyday work will depend less on how visually impressive the output looks than on whether the generated controls are accurate, accessible and easier to use than a well-structured text answer.
