OpenAI adds interactive UI layer to ChatGPT with GPT-6 launch

OpenAI is rolling out "Intelligent UI", a new interface layer that lets ChatGPT embed interactive visual components — charts, forms, buttons — directly in its responses instead of plain text. The update starts today for Plus, Pro, Business and Enterprise subscribers. Go and free-tier users get it Thursday, alongside the release of the GPT-6 Sol and Luna models.

According to a company blog post, GPT-6 was trained to decide when a visual component beats plain text and how to render it inside the conversation flow. One demo shows a seven-speed bicycle gear diagram with buttons that highlight different parts on click. Another teaches Mahjong with visual tiles grouped into scrollable categories. The model can also spin up compact tools — a pension savings calculator, a retro game, a bill splitter — that appear inline.

OpenAI announced the GPT-6 family last month and now splits it into two tiers. Sol, the mid-range model, goes to paying subscribers. Luna, a more compute-efficient variant, serves Go and free users. The split mirrors OpenAI's usual capacity strategy: reserve heavier compute for paying customers while giving the broader audience a lighter version.

Beyond the interface, OpenAI says GPT-6 runs more efficient web searches and can stream partial answers while it gathers more information, a capability sometimes called "insight streaming." The company also claims the model shows "stronger resilience to attempts to bypass its safety training," though it released no concrete benchmarks or testing methodology for independent comparison.

What's missing: no standard benchmarks such as MMLU, HumanEval or GPQA; no context-window size, token pricing or latency figures. The safety-resilience claim remains a vendor assertion without disclosed attack vectors or head-to-head results against prior models. Users will have to wait for community and third-party testing to verify the claimed improvements.