OpenAI / 7 October 2026

ChatGPT Adds Interactive Answers as GPT-6 Reaches Free Users

OpenAI began rolling out GPT-6 with interactive answers in ChatGPT on 7 October, extending the upgrade to Free and Go users from 8 October after its initial paid-tier release. The company’s Intelligent UI can assemble buttons, diagrams, forms and small interactive tools inside an answer, bringing a different interface to a service OpenAI says reaches more than 1.2 billion people weekly. This is a change to the Chat experience, rather than another upgrade to the models used in Work or Codex. Its commercial purpose is readily understandable: more tasks can be completed inside the conversation, increasing the usefulness of ChatGPT as a destination rather than only a source of explanatory text.

The release uses GPT-6 Sol for paid plans and GPT-6 Luna for Free and Go, both adapted for everyday conversation. OpenAI says a component library and compiler let interfaces appear progressively, while the models decide when a visual element helps. Its internal evaluation also found that GPT-6 Instant began answering search questions 44% sooner than GPT-5.6 Instant. That is a time-to-first-answer claim, not evidence that every complete task finishes 44% faster. The distinction is operationally significant: reducing the wait before useful output appears can improve the experience even when the total work, computation and final checking required remain substantial for the user.

An early hands-on account by WIRED’s Reece Rogers described working examples including an apartment affordability calculator and an aircraft-seat explorer. That supplies some independent evidence that the interface can produce usable interactions, although it was a brief preview rather than a systematic reliability assessment. OpenAI product manager Aarush Selvan told WIRED that the team considered when additions become “cluttered.” The choice matters because generated interfaces have two jobs: they must present information accurately and make the controls understandable. A polished calculator is useful only if its assumptions and calculations are sound; visual fluency does not itself establish correctness or suitability for consequential decisions.

The October system card treats the models as having High capability in cyber and biological or chemical domains under OpenAI’s framework, while not reaching its High threshold for AI self-improvement. It reports improved resistance to some attacks alongside regressions on certain safety evaluations, including parts of the under-18 testing. OpenAI describes additional safeguards and review rather than claiming uniform progress on every measure. These are deliberately challenging evaluations, not estimates of how often ordinary users will encounter unsafe answers. Their business relevance is that broader distribution increases the importance of consistent product-level controls, even where the underlying model improves on an average benchmark score.

For software suppliers, the interface change creates a practical distinction between information access and ownership of the customer’s working environment. If ChatGPT supplies an adequate temporary calculator, planner or comparison view, a user may have less reason to open a separate application for that particular task. More durable workflows still require dependable records, permissions and maintenance. That competitive implication is an inference, not a reported displacement of software revenue. Enterprise administrators also retain a role in availability. OpenAI has therefore widened what the consumer interface can express without demonstrating that generated views can replace every controlled business application or the specialist services behind one.