ChatGPT Intelligent UI: GPT-6 Rollout and API Update
ChatGPT Intelligent UI rolls out with GPT-6 Sol in paid Chat; GPT-6 Luna expansion is scheduled next. Check clients, effort limits and the updated chat-latest API alias.
ChatGPT Intelligent UI began rolling out on October 7, 2026, bringing interactive answers to GPT-6 in ordinary Chat. Instead of treating every answer as a block of text, ChatGPT can choose a useful mix of prose, visuals, and controls. The release matters to people comparing options or exploring a calculation in conversation. It also changes which GPT-6 model powers paid Chat, while leaving Work and Codex models unchanged.
What ChatGPT Intelligent UI changes
OpenAI’s announcement describes answers that adapt their presentation to the task: a chart, diagram, form, or button can accompany the explanation. It also describes small tools generated inside a conversation, such as a calculator or bill splitter. Text remains a valid response; the launch is not a promise that every prompt produces a widget.
A native component library and compiler render responses progressively. ChatGPT can start presenting an answer while continuing to think. For web-search questions, OpenAI reports that answers start 44% sooner on average than GPT-5.6 Instant. That measures the beginning of an answer, not the time required to finish every task.
The practical distinction is between reading a conclusion and exploring its assumptions. A useful interactive answer should make those assumptions easier to inspect. An attractive chart with unclear inputs still leaves the original decision unresolved.
Who receives the October 7 rollout
The dated ChatGPT release notes put Plus, Pro, Business, and Enterprise Chat in the rollout starting October 7. Paid Chat is powered by GPT-6 Sol. OpenAI says Free and Go expansion starts October 8 with GPT-6 Luna; on this article’s October 7 verification date, that is an announced next step, not confirmed availability today.
Intelligent UI is supported from Instant through Extra High where those efforts are available to the account. The Pro reasoning option remains powered by GPT-6 Astra and does not support Intelligent UI. A Pro subscription and the Pro reasoning option are therefore different checks.
This follows the earlier Sol and Luna launch, whose September scope covered other product entry points. Today’s ordinary-Chat change should not be read backward into that historical announcement.
Check the client, then the model selection
The current model help page says to use the web or a supported updated app. Older macOS and Windows ChatGPT desktop apps do not support the new capabilities; the web is the alternative. No special prompt is required.
In the model menu, GPT-6 replaces Latest while explicit GPT-5.6 Sol selections remain on GPT-5.6. Check what is selected before comparing two answers. If GPT-6 has not reached an eligible account yet, a missing control alone does not establish an installation fault.
On the web, Settings includes Layout and visuals. Turning it off reduces visual elements but is not a complete classic or text-only mode. Intelligent UI has no separate quota; existing model, tool, and plan allowances apply.
For a first check, record the account plan, client, visible model label, and chosen effort. These four observations help distinguish a rollout or permission question from an answer-format preference.
Business and Enterprise controls still matter
The Business model guide distinguishes Chat usage from Work and Codex usage. A Premium seat is not the same thing as the Pro reasoning option. Business can add reasoning automatically while Instant is selected; Higher intelligence in Settings > General controls that behavior, and automatic reasoning may use the reasoning allowance.
The October 7 Enterprise entry says the rollout uses the existing GPT-6 Sol access setting for Instant through Extra High. Administrators can turn access off, and workspace and role permissions remain applicable. There is no new admin preview period.
For a team pilot, our recommendation is to select one approved task and one supported client, then record the workspace’s model setting alongside the result. Do not instruct staff to switch accounts or bypass a disabled workspace option to reproduce a public demonstration.
Try an inspectable comparison, not a decorative demo
Here is an original, untested example: compare three hypothetical subscription options costing $12, $24, and $40 monthly. Give each option a specific allowance, state the intended monthly usage, and ask ChatGPT to show both the comparison and the assumptions behind its recommendation. Ask for editable inputs if a suitable interactive format is available; that is a request, not a guaranteed output.
Keep a plain-text copy of the three inputs beside the response. Change one price, then check whether the displayed total and recommendation both reflect the change. Ask what happens when usage exceeds an included allowance. Finally, request a short text summary explaining the selected option.
This exercise tests three different things: whether the interaction works, whether the arithmetic is correct, and whether the conclusion follows the stated criteria. A successful button click establishes only the first. Choose a small example that can be checked manually before entrusting a larger decision to the same presentation.
The same-day API update is a moving alias
The October 7 API changelog separately updates the exact model ID chat-latest, pointing to the latest ChatGPT model for Plus, Pro, Business, and Enterprise. OpenAI recommends the GPT-6 family for production API use and describes this alias as a way to test current chat improvements.
Its official model page calls it the latest Instant model used in ChatGPT. It lists a 400,000-token context window, a 128,000-token maximum output, and text-token prices of $5 input, $0.50 cached input, and $30 output per million tokens. The snapshot list supplies the moving alias, not a dated immutable ID.
Do not substitute those numbers for GPT-6 Sol’s API rates or assume an API response reproduces ChatGPT’s interface. Our API model comparison addresses separately named API models. For an evaluation, record the alias, test date, input, output, and cost before rerunning the same cases after an update. That creates a useful comparison even when the underlying snapshot changes.