OpenAI details GPT-6 Intelligent UI availability and limits

OpenAI says the global rollout begins with ChatGPT Plus, Pro, Business and Enterprise users, then expands to Free and Go users starting tomorrow.[1] The paid tiers use GPT-6 Sol in ChatGPT, while Free and Go use GPT-6 Luna; OpenAI says both models are tuned for everyday conversation.[1] The update applies to ChatGPT’s Chat tab, and OpenAI says the models powering Work and Codex are unchanged.[1] OpenAI posted its announcement on October 7, 2026.[1]

OpenAI’s help information says Intelligent UI is rolling out on the web and supported updated apps with GPT-6 at effort levels from Instant through Extra High, but not at Pro effort.[4] It has no separate usage quota, while existing model, tool and plan allowances continue to apply.[4] OpenAI also says the GPT-6 rollout does not retire older selectable models.[4] For teams planning a trial, these details distinguish the new interface’s availability from the broader plan and model allowances already in place.[4]

OpenAI describes Intelligent UI as a capability for quick answers that can make everyday responses more visual, support learning and create task-specific tools.[2] Responses can combine text, visuals and interactive elements according to the question, including graphics, tappable buttons, forms, charts and interactive experiences inside the conversation.[2] OpenAI says a simple text answer remains an option when it is most useful.[2] The company says it built a library of native, streamable components and a compiler that processes interfaces as the model generates them, so interfaces can appear progressively.[2]

OpenAI says GPT-6 can begin answering while it continues to think.[2] The company reports that GPT-6 Instant starts answering web-search questions 44% sooner on average than GPT-5.6 Instant.[2] OpenAI also says GPT-6 Extra High can begin answering in the same time as GPT-5.6 Medium while scoring better overall than GPT-5.6 Extra High.[2] In an internal evaluation, OpenAI says GPT-6 more often addressed the key aspect of difficult questions correctly than GPT-5.6.[2] These are company-reported comparisons; the cited evaluation is described as internal.[2]