OpenAI is rolling out GPT-6 globally in ChatGPT together with a new interactive system called Intelligent UI. Instead of returning text alone, the chatbot can place diagrams, images, buttons, forms, and small tools directly inside an answer. The change affects everyday tasks, learning, and office work, while a separate interface lets organizations automate fast classification decisions.
What GPT-6 changes in ChatGPT
Intelligent UI, an interface that adapts its presentation to the task, is the most visible part of GPT-6. According to OpenAI’s announcement, the model is rolling out globally and is intended to deliver faster responses with visual and interactive elements. One answer can therefore combine explanatory text with a chart, selectable buttons, and a form.
OpenAI says it trained GPT-6 to decide when an interactive presentation will be more useful than plain text. ChatGPT can arrange options side by side for a comparison, create an explorable diagram for an explanation, or provide a small calculator for a numerical task. The interface becomes part of the answer rather than merely its container.
One demonstration features a diagram of a seven-speed bicycle. You can select buttons to highlight individual components and inspect what they do. Another example teaches Mahjong with images of the tiles, organized into categories you can scroll through. OpenAI also showed a savings calculator, a bill-splitting tool, and a small retro game that can be used within the conversation.
GPT-6 comes in several versions. Plus, Pro, Business, and Enterprise accounts initially receive GPT-6 Sol, while Free and Go accounts use the more efficient GPT-6 Luna. The announced global rollout began for paid accounts on October 7, 2026, with Free and Go scheduled to follow the next day. The supplied sources do not state subscription prices.
The update goes beyond presentation. OpenAI says GPT-6 performs better web searches and can start displaying partial results while it continues gathering information. The company claims this reduces waiting time by 44 percent. That figure comes from the provider’s internal reporting and has not been independently verified in the supplied sources; the same applies to the reported accuracy gains over GPT-5.6 and stronger resistance to attempts to bypass safety training.
Why interactive answers matter
Interactive responses can be particularly useful when spatial relationships, processes, or changing values are hard to understand in a block of text. In one presentation, ChatGPT created several diagrams to explain how an airplane wing generates lift. Instead of only reading an abstract description, you can follow the relationships visually.
Everyday planning can also become more concrete. OpenAI’s examples include visual recipes, a map for a multi-day hiking trip, and a savings calculator tailored to the task. For trip planning, you could explore a route, while an editable chart could show how changed inputs affect a financial scenario. The sources establish that these types of tools were demonstrated, but not how reliably they perform in real situations.
The amount of visual material will be customizable. If you prefer concise text responses, you will be able to reduce the visuals, according to the demonstration given to journalists. That control is useful because an elaborate answer is not automatically a better one. A polished calculator can look just as convincing as an accurate calculator, so verification does not disappear with the text box.
The underlying idea is not entirely new. The Decoder reports that Google introduced a similar feature called Neural Expressive in Gemini in May. OpenAI’s distinction therefore lies less in inventing the concept than in broadly integrating it into ChatGPT and combining graphics, forms, editable elements, and tools generated for a particular conversation.
What makes the Decisions API different
Alongside GPT-6 and Intelligent UI, OpenAI introduced the Decisions Application Programming Interface (API), a connection that allows other software to use a service automatically. It is mainly intended for organizations and digital service providers. Even so, it may determine which automated choices you encounter in customer support, document processing, or image assessment systems.
The interface is not designed to write long passages. Instead, it returns a narrowly defined decision about text, images, or a combination of both. It supports three output formats: a probability for a yes-or-no question, a selection from predefined categories, and a score on a specified scale. OpenAI’s examples include detecting damage in photographs, routing customer requests, and classifying documents.
According to the report on the public beta, the Decisions API operates about ten times faster than OpenAI’s Responses API. It currently works only with GPT-6 Luna. The stated price is $0.10 per million input tokens; tokens are small units of text used to calculate consumption. The source says there is no charge for output tokens.
An independent practical note by Simon Willison confirms the three decision formats and support for images. In his example, the model checks a picture for mammals and returns a probability of 0.0. His comparison with the Jev service also shows that OpenAI is not the first provider in this category: Jev charges $0.042 per million input tokens but, according to the source, does not accept images as GPT-6 Luna does.
OpenAI lists support for Zero Data Retention, meaning submitted content is not retained on an ongoing basis, and use under United States Health Insurance Portability and Accountability Act (HIPAA) requirements in the United States and Europe. This is not a blanket assurance that every deployment complies with all Swiss or European privacy requirements. The interface also remains in public beta, so its features, pricing, and terms should not be treated as final.
OpenAI has separately reduced its paid API usage tiers from five to three. The Build, Launch, and Grow levels allow monthly usage limits of $500, $5,000, and $200,000, respectively. Organizations are supposed to move up automatically according to their total credit purchases. These limits do not directly affect ordinary ChatGPT users, but they show that OpenAI is positioning the Decisions API for substantial business workflows as well as small experiments.
Pros and Cons of GPT-6 and Intelligent UI
Pros:
- Clearer explanations – Diagrams and explorable graphics can make technical or spatial relationships easier to understand.
- Tools inside the conversation – You can use calculators, forms, and selection controls without moving between several applications.
- Broad availability – The global rollout covers paid and free ChatGPT accounts, although they receive different GPT-6 versions.
- Fast classification – The Decisions API gives organizations a specialized method for placing images and text into defined categories or rating scales.
Cons:
- Unverified performance claims – Most of the reported speed and quality improvements come from OpenAI’s statements and internal tests.
- Persuasive presentation – A professional-looking graphic or interactive calculator may conceal faulty assumptions more effectively than plain text.
- Different model access – Free accounts receive GPT-6 Luna while higher tiers use GPT-6 Sol, and the sources provide no complete performance comparison.
- Unresolved privacy questions – Zero Data Retention and HIPAA support do not replace an assessment of a specific use, especially when sensitive information is involved.
What this means for you
If you are a beginner, a manageable learning or planning task is a sensible first test. You could ask for an interactive diagram explaining a bicycle’s gears or create a simple tool to divide a shared bill. Afterward, check whether the inputs, labels, and results make sense rather than treating visual polish as evidence of accuracy.
If you are an advanced user, you can specify the type of presentation you need. For a trip, you might request a map with individual stages; for a savings calculation, you could ask for adjustable inputs and an editable chart. If the visual elements become distracting, you can reduce them. The best interface is not necessarily the one with the largest button collection.
For organizations, the distinction between a generative response and a limited decision is central. Routing a customer request to the correct department is a different task from composing a friendly reply to that request. The Decisions API can handle the first step, while critical or ambiguous cases may still require human review. A probability remains less than a certainty, even when it arrives in tidy formatting.
For Switzerland, the global ChatGPT rollout means Intelligent UI is intended to be available to Swiss accounts as well. The sources provide no details about Swiss languages or regional adaptations. For business uses involving personal, health, or customer data, references to Zero Data Retention and HIPAA support do not establish compliance with Swiss requirements; the reports make no definitive claim on that point.
GPT-6 moves ChatGPT from a primarily text-based conversation toward an interface generated for each task. That can make explanations clearer and small jobs more convenient, while the Decisions API offers a narrower and faster form of automated classification. The open issue is how reliably these tools perform in daily use, and whether visual confidence exposes errors or merely gives them better decoration.
Sources
- GPT-6 and Intelligent UI for everyone – OpenAI, 2026-10-07
- ChatGPT’s ‘Intelligent UI’ update fills its responses with pictures, charts, and buttons – The Verge, 2026-10-07
- “Intelligent UI”: OpenAI bringt GPT-6 mit neuer interaktiver Oberfläche in ChatGPT – The Decoder, 2026-10-07
- ChatGPT is getting a lot more visual, with the launch of a new interface – TechCrunch, 2026-10-07
- Neue OpenAI-Schnittstelle trifft schnelle Entscheidungen statt Texte zu generieren – The Decoder, 2026-10-07
- llm-openai-decisions 0.1a0 – Simon Willison’s Weblog, 2026-10-06
Image: Matheus Bertelli via Pexels