GPT-6 comes to everyday ChatGPT with Intelligent UI; Free users follow a day later on Luna

Why it matters

Key points

OpenAI is bringing GPT-6 to everyday conversations in ChatGPT (the Chat tab) together with Intelligent UI: instead of text alone, the model can compose its answer with visuals, tappable buttons, forms, charts and small interactive tools, such as an explorable diagram, a savings calculator, a bill splitter or a game playable inside the conversation. OpenAI says this runs on a library of native, streamable components plus a compiler that renders the interface progressively while the model is still generating.

The second change is answering while thinking: reasoning models no longer have to finish thinking before replying, and can build an answer across several partial responses. OpenAI says that on an internal evaluation of everyday agentic tasks, GPT-6 Extra High starts answering as quickly as GPT-5.6 Medium while scoring better than GPT-5.6 Extra High, and that for questions needing web search GPT-6 Instant starts answering 44% sooner on average than GPT-5.6 Instant.

Rollout: Plus, Pro, Business and Enterprise begin receiving it globally on day one, powered by GPT-6 Sol; Free and Go follow from the next day, powered by GPT-6 Luna. Enterprise availability depends on admin settings. The models behind Work and Codex are unchanged.

Verifiable

Primary page: https://openai.com/index/gpt-6-for-everyone

The OpenAI News RSS dates it October 7, 2026, category Product; the page appeared in OpenAI's news index in the early hours of October 8 Beijing time. Safety document: https://cdn.openai.com/pdf/gpt-6-october.pdf

The speed and score comparisons come from OpenAI's internal evaluations, with no published task set, sample size or third-party replication. The figure of more than 1.2 billion weekly ChatGPT users is also self-reported. On safety, OpenAI says GPT-6 builds on several of Astra's safety advances, holds up better against multi-turn adaptive jailbreaks, and has stronger protections against cyber, biological and violent misuse; see the safety document for details.

Our take

The real shift is not a benchmark score but that the model now decides what an answer looks like. Chatbot output used to be locked into a text box; now the model has to judge whether a question calls for a comparison table, an interactive diagram, or a single sentence. That makes interface design part of model capability, and it means ChatGPT starts to stand in for a class of lightweight tools and small apps: calculators, checklists and trip maps may simply be generated from one request.

Free users getting GPT-6 Luna puts a frontier-generation model in front of a billion-scale audience, which is why this goes into this week's major changes. Two things to watch: OpenAI itself says the model's design judgment still needs work, and a wrong interface is harder to spot than a wrong sentence; and answering while thinking means early partial answers may later be revised, so it remains to be seen whether users can tell what is settled and what is still being reasoned out.