OpenAI begins GPT-6 rollout with Intelligent UI in ChatGPT

OpenAI began rolling out GPT-6 with Intelligent UI in ChatGPT on October 7. The update gives the chatbot a response system that can combine text with visuals, charts, controls, forms and task-specific interactive tools.

The global rollout starts with Plus, Pro, Business and Enterprise accounts in the Chat tab. Free and Go access is scheduled to begin October 8, while Enterprise availability depends on administrator settings. The change applies to ChatGPT’s Chat experience; OpenAI said the models used in ChatGPT Work and Codex are not changing as part of this release.

Intelligent UI chooses how to present an answer based on the prompt. Comparisons can appear side by side, explanations can use interactive diagrams, and tasks can produce interactive tools inside the conversation. ChatGPT can still return plain text when that format is more useful.

OpenAI said the system draws on a library of native, streamable components and a compiler that processes the interface as the model generates it. This allows interface elements to appear progressively, without waiting for the entire response to finish.

The feature is rolling out on the web and supported updated apps for GPT-6 at Instant through Extra High effort. OpenAI’s help page says it is unavailable at Pro effort and in older ChatGPT desktop apps for macOS and Windows.

OpenAI’s launch post says GPT-6 Sol powers eligible paid Chat tiers, while GPT-6 Luna powers Free and Go. A separate OpenAI help article updated the same day says Free and Go use GPT-5.6 Luna instead, leaving the model assignment for those plans unresolved in the company’s documentation.

The company says more than 1.2 billion people use ChatGPT each week. It also says an internal evaluation found that GPT-6 Instant began answering web-search questions 44% sooner on average than GPT-5.6 Instant. Those usage and performance figures come from OpenAI and were not independently validated by the opened sources.

In an accompanying system card, OpenAI said GPT-6 showed stronger resistance to jailbreaks and reduced dishonesty, deception and attempts to bypass guardrails compared with GPT-5.6. The company also disclosed statistically significant regressions in standard evaluations: self-harm for GPT-6 Sol, and self-harm, gore and sexual content for GPT-6 Luna. In dedicated teen-safety evaluations, both models regressed on age-restricted content, sexual content and emotional reliance; GPT-6 Luna also regressed on gore. OpenAI said manual review found the violations generally low severity and described additional classifier-based protections for teen users.

OpenAI classifies the October GPT-6 Sol and GPT-6 Luna releases as High capability in cybersecurity and biological and chemical domains, and below its High threshold for AI self-improvement. These October versions are being introduced in ChatGPT Chat; previously released versions remain in ChatGPT Work and Codex.