Share with your CIO
OpenAI is betting that removing friction for free users converts more of them into paying customers, and the company is restructuring its ChatGPT tiers to test that thesis at scale. Starting next week, free and Go tier users get unlimited text conversations, a “Think” button for on-demand reasoning, and a default upgrade to GPT-5.6 Luna. Paying Plus and Pro subscribers get a meaningfully updated GPT-5.6 Sol model, improved factual reliability on dates and figures, and a slider that lets users dial how much reasoning effort the model applies to any given query.
What this means for your business
The employees who use ChatGPT on personal free accounts are already inside your organization. Unlimited text access with a “Think” toggle is the kind of usability jump that turns casual experimentation into habitual dependency, and habitual dependency is what eventually shows up as a shadow IT conversation or a sanctioned procurement request. Whether you’re the CIO managing an approved AI stack or still building one, this move shrinks the window before someone escalates the question of standardization.
The reasoning slider on Sol is worth noting as a signal, not a feature. OpenAI is surfacing what AI infrastructure teams call compute steering, where the caller decides how much processing power the model actually burns per response. Enterprise pricing models have historically been flat-rate or volume-based. If OpenAI normalizes user-level compute control, it opens a path to usage-tier pricing granular enough to charge by reasoning depth, not just by message. That’s a procurement model your CFO hasn’t modeled yet, and your contract negotiators should know it’s coming.
The factual reliability improvements to Sol specifically targeting dates, numbers, source citations, and rule-based answers are the tell about where enterprise objections are loudest. OpenAI is patching the failure modes that kill procurement approvals in legal, finance, and compliance. If those patches hold up in practice, the “we can’t rely on it for anything high-stakes” objection gets shorter. The falsification condition is straightforward: if accuracy on structured, verifiable queries doesn’t improve measurably in third-party evals over the next quarter, this positioning is marketing copy, not a capability shift.
Based on reporting from OpenAI is giving ChatGPT free users unlimited text chats, originally published 2026-08-06 13:00:00.
