OpenAI is bringing GPT-6 with Intelligent UI to ChatGPT, allowing responses to include interactive diagrams, charts and tools users can operate inside a conversation. The global rollout starts today for Plus, Pro, Business and Enterprise users in the Chat tab, with Free and Go scheduled to follow on October 8. Enterprise access also depends on settings chosen by workplace administrators.
The paid tiers will use GPT-6 Sol, while Free and Go will receive GPT-6 Luna, with both models adapted for everyday conversations. The update covers Chat in ChatGPT; the models behind Work and Codex remain unchanged. OpenAI says the new model is intended for ChatGPT’s audience of more than 1.2 billion weekly users, following the first paid-customer GPT-6 introductions last month.
Read more: OpenAI to Add Text Watermarking for EU ChatGPT and Codex Users
OpenAI says Intelligent UI lets GPT-6 choose a response format to suit the question, including a plain text answer when that is sufficient. Users can ask for a savings calculator, explore an explanation by changing inputs, or have travel stops plotted on a map. Responses can also incorporate forms and buttons that users tap, with the experience designed for both mobile and web.
OpenAI built a component library and a compiler that renders interfaces during generation, so users can see portions before the response is finished. The company says it trained GPT-6 to arrange information and decide when interaction is useful, evaluating generated interfaces for how clearly and completely they address a task. OpenAI acknowledges that the model still needs better judgment about design and a broader range of things it can create.
GPT-6 can also start responding before its reasoning has finished, adding information through successive partial answers. For requests involving web searches, OpenAI says GPT-6 Instant begins responding an average of 44% sooner than GPT-5.6 Instant. That comparison measures the wait before an answer starts, rather than the time needed to finish the response.
Earlier this month, GPT-6 Astra Ultrafast became available through the OpenAI API and reached eligible users of ChatGPT Work and Codex, running on NVIDIA Blackwell GPUs. GPT-6 Sol and GPT-6 Luna had also reached GitHub Copilot in September, where both carried usage-based billing and availability varied by subscription. For the new ChatGPT release, OpenAI says GPT-6 safety training draws on real-world use to strengthen protections against misuse involving cyberattacks, biological threats and violence.
