OpenAI has started putting its GPT-6 artificial intelligence models into ChatGPT for everyone, including people on the free plan, and the change brings a new way of answering that will look different to the many New Zealanders who use the chatbot.
In an announcement dated 7 October, OpenAI said GPT-6 was rolling out globally to its paid Plus, Pro, Business and Enterprise plans that day, with the free and Go plans following from the next day. Paying subscribers get a model called GPT-6 Sol. Free and Go users get GPT-6 Luna. OpenAI says both are tuned for everyday conversation.
The company had already released the first GPT-6 models to paying customers last month. This week’s release replaces the GPT-5.6 models in the main Chat tab of ChatGPT, rather than its specialist work and coding tools.
The headline feature is what OpenAI calls Intelligent UI. UI stands for user interface, the buttons, boxes and screens people tap and click. OpenAI describes Intelligent UI as a capability that lets ChatGPT answer with fully interactive screens, making everyday answers more visual.
In practice, OpenAI says an answer can now include graphics, tappable buttons, forms, charts and small interactive tools inside the conversation, instead of only paragraphs of text. Its examples include putting the stops of a road trip on a map, laying out cooking times beside a recipe, and building a bill splitter or a savings calculator on request. OpenAI says the model decides which format suits each question, and can still give a plain text answer when that is more useful.
Behind the scenes, OpenAI says it built a library of ready-made screen components and trained GPT-6 to choose and arrange them, so the interface appears piece by piece as the answer is written. The company calls Intelligent UI a significant shift in how people use ChatGPT, while also acknowledging there is still work to do on the model’s design judgement.
The other change is speed. Newer “reasoning” models, which work through a problem before replying, have tended to make people wait. OpenAI says GPT-6 can now start answering while it keeps thinking. It says that for questions needing a web search, GPT-6 Instant starts answering 44% sooner, on average, than the previous GPT-5.6 Instant. That figure comes from OpenAI’s own testing, not from an independent evaluation.
OpenAI describes the rollout as global, so it applies to New Zealand accounts too. On its pricing page, the free plan is now listed as offering unlimited text chats with GPT-6, subject to what OpenAI calls abuse guardrails, with limits on uploads, image creation, voice and other tools. The same page notes that the Go plan may include ads.
The update applies only to the Chat experience. OpenAI says the models behind its Work and Codex products are not changing as part of this release. For businesses on Enterprise plans, availability depends on the settings chosen by the organisation’s own administrators.
The release reaches a large local audience. In September, OpenAI told 1News that New Zealanders were sending almost 10 million messages to ChatGPT each day, and that a survey it commissioned found 40% of respondents used ChatGPT every day. OpenAI also said more than a third of ChatGPT messages sent in New Zealand were work related, about six percentage points above the global average. “Weekly active users in New Zealand have grown by more than 20% over the past year,” OpenAI’s head of communications for New Zealand and Australia, Chloe Browne, told 1News. These are the company’s own figures.
Alongside the launch, OpenAI published a system card, a technical report on how the models performed in its safety testing. It gives a more mixed picture than the launch post.
OpenAI says that, compared with the GPT-5.6 models they replace in ChatGPT, the new GPT-6 models showed stronger resistance to jailbreaks, which are attempts to trick a model into ignoring its safety rules, as well as less dishonesty and deception. It also classes the October release of GPT-6 Sol and GPT-6 Luna as “High capability” in both cybersecurity and biological and chemical risk under its internal Preparedness Framework, and says it has applied the same safeguards it used for GPT-5.6.
The report also lists areas where the new models did worse on OpenAI’s own deliberately difficult test sets. It says GPT-6 Sol showed a statistically significant regression on self-harm content, and GPT-6 Luna, the free-plan model, showed regressions on self-harm, gore and sexual content. On tests designed for users under 18, both models showed statistically significant regressions on age-restricted content, sexual content and emotional reliance, a category covering emotional dependence on the chatbot.
OpenAI says it manually reviewed the failures and found that “the disallowed responses were generally low severity”. It says the models are more willing to answer informational questions about self-harm while pointing people to professional help, and that it applies an extra filter to responses for younger users that is not reflected in the test scores. On the emotional reliance result, the company says its test is overly sensitive to the model using friendly nicknames such as “bro” or “bestie”. OpenAI also cautions that these tests focus on hard cases and should not be read as estimates of how often such behaviour happens in normal use.
For parents and schools in New Zealand, the practical point is that the model on the free plan is GPT-6 Luna, and OpenAI’s own report shows several statistically significant regressions for it on the under-18 tests.
OpenAI frames the release as the start of a bigger change in how people use software, writing that it expects software to adapt to people rather than people having to learn fixed interfaces. Whether interactive answers prove more useful than text, and how well they handle New Zealand-specific questions, will become clearer as more people use them.
Have you noticed the new interactive answers in ChatGPT yet, and are they more useful than plain text? Share your experience in the comments below.
This article was written by AI, briefed to report the facts, hopefully without some of the bias people bring to the job 🙂

Leave a Reply