OpenAI Rolls Out GPT-6 with Interactive UI to 1.2 Billion ChatGPT Users

OpenAI has begun a global rollout of GPT-6 paired with Intelligent UI, a new feature that enables ChatGPT to generate interactive elements—including charts, buttons, forms, calculators, and diagrams—directly within conversations. The feature allows users to adjust variables in scenarios such as budget planning or travel itineraries without rephrasing questions, and ChatGPT can begin answering while still processing or reasoning through a problem.

The rollout follows a tiered schedule. Paid subscribers—those on Plus, Pro, Business, and Enterprise plans—gained access to GPT-6 Sol with Intelligent UI on October 7, 2026. Free and Go users began receiving GPT-6 Luna on October 8. OpenAI stated that more than 1.2 billion people use ChatGPT weekly and are eligible for the update, which applies only to the Chat experience; Work and Codex continue to use their existing models.

Intelligent UI represents a shift in how ChatGPT responds to queries. Rather than delivering all answers in text format, GPT-6 can choose the most useful presentation method—visual explanations for conceptual topics, side-by-side comparisons for evaluations, interactive diagrams for technical processes, and embedded tools for immediate tasks. Users can turn off visual layout in web settings if preferred, though some visual elements may still appear. The feature is available at Instant through Extra High thinking levels on web and supported updated apps, but not in older macOS and Windows desktop clients or at Pro thinking effort.

OpenAI reported performance improvements alongside the release. For questions requiring web search, GPT-6 Instant begins answering 44% sooner on average than GPT-5.6 Instant, though this measures response start time rather than completion time. GPT-6 also improved in deciding when to search and finding supporting information. The company built the feature using native, streamable components and a compiler that displays interfaces progressively as the model generates responses, avoiding delays for complete generation.

Alongside the rollout, OpenAI published a deployment safety report for GPT-6 Sol and Luna. The report noted that both models achieved High (but not Critical) ratings in cybersecurity and biochemistry domains, with improvements in resisting multi-turn jailbreak attacks and instruction-level tests—Luna and Sol reached 99.79% and 99.99% resistance to prompt injection, respectively. Factual accuracy improved across medical, legal, and financial domains. However, the report also documented some performance regressions: both versions showed declines in filtering self-harm, graphic violence, and sexual content compared to GPT-5.6, particularly for minors. OpenAI attributed this partly to the models' increased willingness to answer informational questions about sensitive topics and noted ongoing deployment-level safeguards. The company acknowledged that evaluation samples were curated for difficulty and may not reflect typical user interactions.