Announced 7 Oct 2026 · Sources checked
What did OpenAI announce today?
OpenAI dated the post 7 October 2026. The same-day RSS item uses the title “GPT-6 and Intelligent UI for everyone” and describes a global ChatGPT rollout with visuals you can use in the conversation. openai.com HTML returned 403 to our fetcher; the sentences below come from official prose on the Wayback Machine snapshot captured at 18:01 UTC that day, plus that RSS item.
The opening paragraph says OpenAI introduced the first GPT-6 models for paid customers last month and is now bringing “that next generation of intelligence” to more people, with a new GPT-6 model “built for more than 1.2 billion people who use ChatGPT each week.” That weekly-user figure is OpenAI’s, and the page repeats it in the closing. It is not an independently audited metric.
This is a follow-up, not a second launch of the September models. On 22 September OpenAI’s RSS listed “Introducing GPT-6 Sol and Luna” as two models for everyday work with different balances of capability and cost. On 29 September the company posted GPT-6.1 Sol for the API. Today’s page puts GPT-6 Sol and GPT-6 Luna into ChatGPT’s Chat tab with a new interface layer. API rates sit in our GPT-6.1 Sol cost and capability note; they are not restated here.
What is Intelligent UI supposed to do?
OpenAI defines Intelligent UI as a ChatGPT capability that “allows ChatGPT to answer quickly with fully interactive user interfaces.” The stated jobs are more visual everyday answers, easier explanations of complex topics, and the ability to “quickly create a tool to solve a task right in the moment.”
The company says it trained GPT-6 to compose replies from text, visuals and interactive elements. Listed building blocks are graphics, tappable buttons, forms, charts, and interactive experiences in the conversation. A comparison might be side by side; an explanation might use a diagram; a simple text answer is still allowed when that is “the most useful response.”
The page’s examples are product illustrations, not tests we ran: a Sunday roast plan, road-trip stops on a map, a seven-speed bicycle broken into systems, a bill splitter, a retirement-savings calculator and a retro game. Topic chips include the central limit theorem, GDP and the Monty Hall problem. Those labels show intended use, not measured lesson outcomes.
How does OpenAI say the interface is built?
The technical description is short. OpenAI says it built “a library of native, streamable components, along with a compiler that processes the interface as the model generates it.” The library is meant to keep a familiar design. The model still decides how the pieces fit. The compiler is what lets the interface appear progressively “without waiting for the entire response to be complete.”
Training, in OpenAI’s account, was expanded so the model can decide content, layout, visuals and interaction. It says it evaluated generated interfaces for clarity, usefulness and completeness, including when text is enough, and that there is “still work ahead to improve the model’s design judgment.” That is an official caveat, not an independent design review.
A separate section says ChatGPT can give an initial response, then think further and follow up. OpenAI contrasts that with earlier reasoning models that waited until thinking finished, and describes interleaving thinking with answering while keeping partial replies cohesive. We have not watched a streamed Intelligent UI reply.
Who gets GPT-6, and which model?
“GPT-6 with Intelligent UI starts rolling out globally to ChatGPT Plus, Pro, Business and Enterprise tiers today in the Chat tab. Starting tomorrow, the rollout expands to Free and Go tiers.” On a 7 October announcement, “tomorrow” is 8 October. “Enterprise availability depends on workplace admin settings.”
“GPT-6 in ChatGPT is powered by GPT-6 Sol for Plus, Pro, Business, and Enterprise tiers, and GPT-6 Luna for Free and Go tiers. Both are tuned for everyday conversation.” Luna also appears on a different surface: the 6 October Decisions API beta accepts only gpt-6-luna for typed predicates, choices and scores. That endpoint is not this Chat-tab UI. See OpenAI’s Decisions API beta.
The team limit is explicit: “This update applies to the Chat experience in ChatGPT. The models powering Work and Codex are not changing as part of this release.” A Work project or Codex session is not covered by today’s wording.
| ChatGPT tier | Model named for Chat | OpenAI’s start |
|---|---|---|
| Plus, Pro, Business, Enterprise | GPT-6 Sol | 7 October 2026, Chat tab; Enterprise needs admin settings |
| Free and Go | GPT-6 Luna | 8 October 2026, Chat tab |
| Work and Codex | Unchanged in this release | Not part of this update |
What do the speed and search claims actually say?
OpenAI reports two timed comparisons and one quality comparison. All three are labelled as internal evaluations. None names a public harness, a sample size, or a third-party auditor.
On “high-value, everyday agentic tasks,” “GPT-6 Extra High is able to begin answering in the same amount of time as GPT-5.6 Medium while achieving a better overall score than GPT-5.6 Extra High.” That is start-of-answer latency plus an unnamed overall score. It does not say every ChatGPT user is on Extra High.
For questions that need web search, “GPT-6 Instant starts answering 44% sooner, on average, than GPT-5.6 Instant.” OpenAI also says GPT-6 chooses more carefully when to look something up and, on an internal set of difficult problems, “correctly addressed the key aspect of the user’s question more often than GPT-5.6.” Those are vendor judgements. For a method that does not take the vendor’s word, see how to verify AI answers.
What remains unproven?
The safety section says GPT-6 builds on “several of Astra’s safety advances,” resists bypass attempts more strongly (especially multi-turn adaptive attacks), and is clearer about its limits relative to GPT-5.6 Sol. OpenAI says it updated training against high-risk misuse involving cyberattacks, biological threats and violence, and that the model should refuse fewer harmless requests. Alignment evaluations during training are said to have looked at honesty and safety restrictions.
The only pointer for numbers is “More detail on the results and our review is available in the system card.” The page we inspected did not link a new card there. We did not open a 7 October GPT-6 Sol/Luna Chat system card, so we are not restating Astra’s 3 September card as if it covered this release.
We did not generate an Intelligent UI bicycle or calculator, and we did not compare Sol with Luna on the same prompt. For recurring work, the useful test is still the task you actually run, as in our assistant comparison worksheet. Today’s news is that OpenAI turned on a Chat-tab interface compiler and named which GPT-6 variant each consumer tier is supposed to get.
Common questions
Is this the same release as GPT-6.1 Sol?
No. GPT-6.1 Sol was an API post dated 29 September 2026. Today’s ChatGPT page names GPT-6 Sol for paid Chat tiers and GPT-6 Luna for Free and Go.
Do Free users get GPT-6 Sol?
Not according to the 7 October post. Free and Go are listed on GPT-6 Luna, with that rollout starting 8 October in the Chat tab.
Are Work and Codex getting Intelligent UI today?
OpenAI says this update applies to the Chat experience and that the models powering Work and Codex are not changing as part of this release.
What to remember
OpenAI is putting GPT-6 Sol and Luna into ChatGPT’s Chat tab and attaching a streamed component library it calls Intelligent UI. Check your plan and admin settings before you assume you have it, keep Work and Codex on their current models until OpenAI says otherwise, and treat the 44% Instant and Extra High figures as internal vendor results.
Sources & further reading
How this story was made
Written by Kristian Kostov with AI assistance and checked against the linked sources. Company performance claims are attributed to the company. Analysis reflects AiLookout’s interpretation; we have not independently tested the products discussed. Cover photography is illustrative and does not depict the specific announcement or product.
Our editorial standards
