OpenAI is adding a Premium seat to ChatGPT Business, priced at $125 per user per month. The tier targets heavy agent users with five times more usage and the removal of the five-hour cap.
Key Takeaways
- Premium seat at $125 per user per month, or $100 with annual billing, against $25 for the Standard seat
- Five times more usage than Standard and no more five-hour cap, inside the same secure workspace
- Up to $500 in workspace credits for early sign-ups, with the waitlist open until August 20
Have an AI Sum Up This Article
ChatGPTFive times the usage for the heaviest profiles
OpenAI is segmenting its team offer. ChatGPT Business gains a Premium seat at $125 per user per month, dropping to $100 with annual billing. The Standard seat stays at $25 per month, or $20 billed annually. The company lays out the details in its official announcement published this week.
The seat’s substance fits in two promises. Premium users get five times more usage than Standard, and the five-hour usage cap disappears entirely. OpenAI frames it as room for larger projects with fewer interruptions, inside the same secure Business workspace.
The ratio between the two tiers is worth stating plainly. Premium costs exactly five times the Standard seat and delivers exactly five times the usage, on monthly and annual billing alike. The grid is linear, which leaves little room to argue about value: buyers are paying for load, not for a list of extra features.
That linearity is itself a positioning choice. OpenAI could have bundled exclusive capabilities into the higher tier and charged a premium on scarcity, the way most enterprise software does. Selling raw headroom instead signals that capacity, not functionality, is the constraint customers are actually hitting.
Both seat types coexist in a single workspace. An admin can reserve Premium for the intensive profiles and keep everyone else on Standard, without migrating anyone or changing the overall plan.
Keeping both tiers inside one workspace matters more than it sounds. Enterprise buyers have learned to distrust upgrades that fragment their tooling, because a split forces two sets of permissions, two audit trails and two support paths. Here the security perimeter stays intact and only the meter changes.
OpenAI is priming the pump with credits. Eligible workspace owners who sign up early receive $100 worth of credits per Premium seat added, or 2,500 credits, capped at $500 for five seats. The waitlist runs until August 20, ahead of early access.
Agents are what broke the old usage math
A heavy-usage tier does not appear out of nowhere. Agentic workloads structurally burn more tokens than a chat exchange: an agent plans, re-runs searches, rewrites and verifies where a chatbot returns a single answer. Quotas designed for conversation crack under that load.
The product calendar tells the same story. Since July, OpenAI has been pushing its GPT-5.6 family toward longer autonomous tasks, and we covered the GPT-5.6 Luna price repositioning that reshuffled the consumer plans. The Premium seat extends that logic to the enterprise side.
The five-hour cap is the detail that gives the tier away. A limit expressed in time rather than in messages only makes sense once sessions run long and unattended, which is precisely how agent work behaves. Removing it is an admission that the previous ceiling was built for a different product.
The target audience is narrow but lucrative. In most teams, a minority of users concentrates most of the consumption: analysts, developers, data-heavy roles. Charging that minority five times more avoids raising the price of every other seat in the building.
There is a defensive read on the same move. Heavy users who keep hitting a ceiling are the ones most likely to test a competing model, because friction is what pushes an engineer to open a second tab. Selling them headroom keeps that experiment from ever starting.
The trade-off echoes telecom pricing. A single flat plan subsidizes heavy users with the margin of light ones, until the gap becomes untenable. OpenAI is moving to intensity-based pricing before inference costs force its hand.
More articles on Horizon
- Claude Watermark Now Covers All Generated Text
- DiffusionGemma Writes Text Four Times Faster
- GPT-5.6 Cyber Writes Attack Code for Defenders
Enterprise AI budgets now get managed per head
For IT leaders, the impact is immediate: AI spend becomes an allocation exercise. Identify the genuinely intensive profiles, justify a five-fold cost per head, measure the return. Adoption documented all the way into ChatGPT’s dominance of AI purchasing in the US Congress shows these decisions long ago stopped being a tech-sector question.
Measuring that return is where most teams will struggle. Hours saved on a drafting task are easy to feel and hard to book, and few organizations instrumented their workflows before handing them to a model. The seat forces a question that adoption alone let everyone postpone.
Well-equipped teams gain real predictability. The five-hour cap used to interrupt long working sessions at the worst possible moment; removing it concretely changes the day for users who chain agent tasks, the same crowd that watched Workspace Agents replace Custom GPTs across their spaces.
Finance teams inherit a harder conversation than IT. A seat that costs five times more per head needs a named owner and a measurable output, which pushes AI spending out of the general software envelope and into per-project justification. Renewal season will be the first real test of that arithmetic.
Smaller companies feel the split differently. A ten-person team where two people run agents all day now faces a bill shaped like a much larger organization, without the procurement leverage that usually comes with it. Intensity pricing does not scale down as gracefully as flat pricing did.
On the competitive front, the reading grid shifts. The question is no longer the price of a seat but the price of intensity. Anthropic and Google will need their own answer for teams whose agents run all day, or watch those accounts drift.
There is a quieter signal in the credits mechanic. OpenAI is effectively paying workspaces to reveal who their heavy users are, seat by seat. That usage map is worth more than the $500 giveaway: it shows exactly where agentic work has taken root inside each company.
Watch the annual billing line for the real intent. Dropping the Premium seat to 100 dollars on a yearly commitment buys OpenAI something more valuable than the discount costs: visibility on committed revenue at the exact moment its infrastructure bills come due. Locked-in seats plan a data center better than monthly churn does.
The move also sketches the market’s next phase. After land-grab pricing comes monetization of real load: every vendor will eventually charge agentic usage at what it costs. The $125 seat is the first visible brick of that shift, and it will not be the last.
Follow the story on Horizon.
