Above: The Mission Bay, San Francisco building that housed OpenAI's headquarters when this photo was taken in June 2025. OpenAI announced GPT-6 for ChatGPT on 7 October 2026. (Photo: Coolcaesar, CC BY 4.0 (Wikimedia Commons))
OpenAI announced on 7 October that GPT-6 is coming to everyone in ChatGPT, on free and paid plans worldwide, according to the company's system card for the release. Alongside it comes Intelligent UI, a feature that lets ChatGPT answer with what OpenAI calls "fully interactive user interfaces", as quoted by MacRumors. Everything below about capabilities is OpenAI's own description, as reported by MacRumors and Search Engine Journal (SEJ) or set out in its system card, unless stated otherwise.
Who gets what, and when
Which model you get depends on your plan, according to MacRumors and SEJ:
- Plus, Pro, Business and Enterprise users get GPT-6 Sol. MacRumors reported the rollout to these plans as happening on 7 October.
- Free and Go users get GPT-6 Luna. MacRumors reported that the rollout would expand to these tiers the next day, 8 October.
SEJ adds that Enterprise access varies with admin configuration, and that Intelligent UI is being introduced gradually on the web and in supported, updated apps. These are announced rollout dates, not confirmation that every account has the feature.
The new models replace GPT-5.6 Sol and GPT-5.6 Luna in ChatGPT, the system card says. MacRumors reports that GPT-6 is limited to the Chat tab. ChatGPT Work and Codex are not changing: the system card says users of GPT-6 Sol and Luna in Codex and via ChatGPT Work are still using previously released versions. To keep the two apart, the card labels the new releases "October" and the ones still in use in Codex and Work "September".
SEJ says Sol and Luna first launched on 22 September for ChatGPT Work, Codex and the API, when OpenAI said they were not yet available in Chat.
What Intelligent UI does
According to SEJ, OpenAI says it trained GPT-6 to decide how each response is laid out based on the question. MacRumors reports that responses will include graphics, tappable buttons, forms, charts, diagrams and other interactive experiences. SEJ notes that ChatGPT can still answer in plain text when that is the most useful format.
MacRumors describes sample comparisons between GPT-5.6 and GPT-6 on OpenAI's website. They show more visuals in responses, an interactive recipe widget that adjusts quantities by guest count, checklists, integrated maps, and multiple-choice follow-up questions to narrow down results. These are OpenAI's chosen examples.
ChatGPT can also build small tools on request. SEJ quotes the announcement: "You can also ask ChatGPT to create a tool for the task at hand, such as a calculator to explore how savings could grow, a bill splitter for dinner with friends, or a game you can play in the conversation."
On the technical side, MacRumors reports that, according to OpenAI, Intelligent UI uses a library of native, streamable components with a compiler that processes the interface as the model generates it.
SEJ lists the limits. The feature is available at thinking levels from Instant to Extra High. The Pro thinking level continues to run on GPT-6 Astra and does not support it, and neither do the older ChatGPT desktop apps for macOS and Windows. On the web, users can turn off a "Layout and visuals" setting, though SEJ, citing OpenAI's help page, says some visual elements might still appear.
Speed and search claims
MacRumors and SEJ both report that GPT-6 can start answering while it is still thinking. SEJ adds that it can share more without the user asking again. For questions that need web search, SEJ reports that OpenAI says GPT-6 Instant starts answering 44% sooner on average than GPT-5.6 Instant. SEJ also reports OpenAI saying GPT-6 makes better decisions about when to look something up. In an internal test of difficult problems, SEJ says OpenAI reported GPT-6 addressed the key aspect of the user's question more often than GPT-5.6, without giving a figure.
SEJ points out one gap: the announcement does not say how sources will appear when responses are presented as charts, comparisons or tools.
What the system card says about safety
The system card is OpenAI's own assessment. It says GPT-6 in ChatGPT incorporates Astra's safety advances, and that compared with GPT-5.6 Sol and Luna in ChatGPT it showed stronger resistance to jailbreaks, including multi-turn attacks, and reductions in dishonesty, deception and circumvention of guardrails.
It also reports regressions. Relative to their GPT-5.6 counterparts:
- GPT-6 Sol (October) shows a statistically significant regression on standard self-harm.
- GPT-6 Luna (October) shows statistically significant regressions on standard self-harm, gore and sexual content.
- On OpenAI's evaluations for users under 18, both models show statistically significant regressions on age-restricted content, sexual content and emotional reliance, and Luna also on gore.
For the main benchmark regressions, OpenAI says it manually reviewed the violating responses and found them borderline but still generally safe, and that adversarial red-teaming did not identify any high-severity risks based on its evaluation criteria. On the under-18 emotional-reliance result, OpenAI says its evaluation is overly sensitive to the model's use of benign nicknames such as "bro" or "bestie". It also says these evaluations were built from difficult cases and are not representative of average production traffic.
Under its Preparedness Framework, OpenAI says it is treating this release of GPT-6 Sol and Luna as High capability in both Cybersecurity and Biological and Chemical domains. It says neither reaches its High threshold in AI Self-Improvement, and that it has applied the same safeguards as those detailed in the GPT-5.6 System Card.
Why it matters
Intelligent UI changes what a chatbot answer can be: a form or a calculator instead of a paragraph. That is my reading, and it is only as good as the examples OpenAI has chosen to show. The 44% speed figure and the search-quality claims rest on OpenAI's statements as reported by SEJ, the safety comparisons come from OpenAI's own system card, and MacRumors's account of the sample comparisons likewise rests on OpenAI's own website.
Sources
Clover, J. (2026) 'ChatGPT Gets GPT-6 and New Intelligent UI With Visual Answers', MacRumors, 7 October 2026, available at: https://www.macrumors.com/2026/10/07/chatgpt-intelligent-ui/ (Accessed: 8 October 2026).
OpenAI (2026) 'GPT-6 Sol and GPT-6 Luna: October 2026 update' [System card], OpenAI, 7 October 2026, available at: https://cdn.openai.com/pdf/gpt-6-october.pdf (Accessed: 8 October 2026).
Southern, M.G. (2026) 'ChatGPT Gets GPT-6 And Intelligent UI For Interactive Answers', Search Engine Journal, 7 October 2026, available at: https://www.searchenginejournal.com/chatgpt-gpt-6-intelligent-ui/592249/ (Accessed: 8 October 2026).