ChatGPT ships GPT-6 and an intelligent UI: what it changes now

On October 7, 2026, OpenAI started rolling out GPT-6 Sol to ChatGPT’s paid plans (Plus, Pro, Business and Enterprise) and, a day later, GPT-6 Luna to the free Free and Go tiers, alongside a feature that changes how the chat gets used at work: Intelligent UI, an interface that no longer answers only in text but generates buttons, tables, calculators and charts on the fly, depending on what’s being asked (Moncloa; also confirmed by WWWhatsnew). At Evicron, an AI and custom software studio based in Barcelona, this is the question more and more clients bring us if they already use ChatGPT daily: if the chat can now “build” a table or a dashboard on the spot, why commission a custom internal tool at all? The honest answer has real nuance.

What exactly changed in ChatGPT

GPT-6 ships in two variants, as we already flagged when Opus 5.5 and GPT-6 Sol/Luna launched back in September: Sol, the full version for paid plans, and Luna, a lighter one for free plans. The real change for daily work isn’t the model itself, it’s Intelligent UI: instead of returning a text block or a Markdown table, ChatGPT can now generate an interactive widget inside the same conversation — a calculator whose fields recompute themselves, a sortable table, a bar chart, a form with buttons. On Enterprise accounts, it’s the admin who decides whether this interface gets switched on for the whole organization or stays off while it’s being evaluated. Pro’s reasoning mode still runs on GPT-6 Astra for now, without this interface, pending a future rollout.

Where this already does the job of an internal tool

Some tasks are now reasonably covered by this feature, work that used to mean opening a spreadsheet or commissioning a small build:

• One-off calculations with several variables — simulating the cost of a new hire, comparing two vendor quotes, a quick ROI estimate. It used to be a spreadsheet formula; now it’s a question answered with an interactive calculator.
• Ad hoc reporting on data that’s already in the conversation — pasting a list of figures and asking for the matching chart, with no other tool involved.
• Interface prototypes to validate an idea before commissioning anything: seeing roughly how a form or a dashboard would look helps specify what’s actually needed.

For a one-off need, or for a single person, this is software that used to cost money and is now included in a subscription the company already pays for.

Where it doesn’t replace custom software

The limit shows up the moment that tool has to stop being a conversation and become a system:

• Real data persistence. A widget inside a chat doesn’t keep a queryable history, version control or backups — every conversation is its own island.
• Several users at once, with different permissions. A dashboard that must show different data depending on who’s looking at it — a salesperson, a manager, a client — needs roles and access control that a single chat thread doesn’t offer.
• Integration with the systems the business already runs. An ERP, a CRM or a customer database doesn’t connect to a ChatGPT conversation without a specific, auditable development layer.
• Deterministic reliability. A financial report or a payroll calculation can’t depend on the model “reading the question right” every single time; it needs fixed, tested, versioned logic.
• Ownership of the data and the code. What gets built inside a conversation isn’t a company asset that another team can audit, modify or migrate to a different provider without friction.

We covered this same boundary, from a different angle, when we looked at whether vibe coding and no-code tools replace custom software: the conclusion holds here too. An interface generated on the fly handles the one-off and the exploratory well; it doesn’t handle what still needs to work the same way in six months, with a different team, under an audit.

What to do before rolling it out to the whole team

1. If you manage an Enterprise or Business account, pilot Intelligent UI with a small team before switching it on for the whole organization. It’s a new feature, worth testing first against your actual business questions.
2. Draw the line between a one-off question and a process that repeats every week. The second one is a candidate for becoming its own tool, not a conversation someone has to retype each time.
3. Don’t paste client data or sensitive figures into a prompt just because the answer “looks like” an app: it’s still a conversation with an outside vendor, under the same privacy rules as any other AI use at the company.
4. If one of these improvised tools has already become essential, that’s the sign it deserves a real build, with the data and the logic kept where the company can actually control them.

In our AI and custom software consulting, that’s exactly where we start: looking at which part of what’s already solved with a chat deserves to stay that way, and which part has grown enough to become its own tool, without depending on a vendor changing its interface again.

In short

GPT-6 Sol and Luna, with their new Intelligent UI, turn ChatGPT into something closer to an app than a chat: it now generates interactive calculators, tables and charts on the spot, at no extra cost over the existing subscription. For one-off calculations, ad hoc reports and prototypes, it genuinely replaces what used to be a spreadsheet or a small build. For anything that needs to persist, serve several users with different permissions, integrate with existing systems or answer deterministically, it’s still custom software’s territory.

Not sure whether what your team is doing in ChatGPT should become its own tool? Get in touch: the first consultation is free and we reply within 24 hours.