Pricing the Unpriced: What Additional Dots Probably Cost

Here is everything OpenAI has actually disclosed about dots pricing as of September 30, 2026: the first dot is included at no extra charge for ChatGPT Pro and Business Premium subscribers in eligible markets. That's it. What a second dot costs, what running several "specialist" dots as a team costs, and what Enterprise, Edu, or Healthcare customers will eventually pay once their admin-enabled beta graduates to general availability — none of it appears in the launch coverage reviewed for this piece. OpenAI hasn't disclosed it. That's not an oversight worth glossing over; it's a deliberate choice at launch, and it's worth taking seriously as a choice rather than filling the gap with a number that sounds plausible.

What "the first one's free" is actually doing

Giving away the first dot is a classic freemium wedge, and it works the same way it always has: let the user experience the thing that's genuinely differentiated — an agent that keeps working after you close the tab — before asking them to pay incrementally for more of it. The bet is that once someone has watched a dot proactively research something useful in the background, or return a pull request they didn't ask for that morning, the case for a second dot makes itself. OpenAI doesn't need to price that second dot aggressively on day one; it needs the first one to be good enough that the pricing conversation happens at all.

That's a sound strategy for OpenAI and a genuinely awkward one for anyone trying to budget for dots today. A team lead who wants to run a dot per function — one watching engineering feedback, one on sales pipeline comparisons, one keeping product docs current — has no way to model the actual cost of that setup right now. They can try it for free, once, and extrapolate nothing reliable about the price of scaling it.

What a reasonable guess looks like, clearly labeled as a guess

Because dots run on dedicated cloud compute — each one gets its own cloud computer and browser, separate from the user's own machine — the cost structure almost certainly resembles compute-metered SaaS pricing more than flat per-seat pricing. A dot that runs continuously, doing proactive research around the clock, consumes real infrastructure whether or not a human ever looks at the output. That's structurally closer to a cloud VM bill than to an extra seat on a chat product, and pricing built to reflect that would likely scale with usage — how much a dot does, not just how many dots exist — rather than a flat per-dot subscription fee.

That's an inference from the architecture OpenAI has described, not a number OpenAI has stated. It would be a mistake to read it as anything firmer than "here's the shape a sensible pricing model would probably take, given what dots actually costs to run" — and readers should treat any specific dollar figure circulating online right now with the same skepticism, since no such figure traces to an OpenAI source as of this writing.

The gap between "first dot free" and "everything after that: TBD" is where the real economics of always-on agents will actually get tested — and OpenAI is, for now, letting usage data accumulate before it commits to a number.

Specialist dots are an even bigger unknown

The pricing gap widens further once you look past the consumer-facing free dot to what OpenAI calls specialist dots — agents with organizational identity and credentials, handling things like procurement, invoicing, or contracting inside a business. These are described as being in early pilots with direct engineer support, and commercial pricing for them is explicitly not disclosed anywhere in the materials reviewed. That's a meaningfully different product from a Pro subscriber's free dot — closer to a managed enterprise deployment than a subscription add-on — and enterprise pricing for that class of product typically gets negotiated per-account rather than published on a pricing page at all. Anyone budgeting for specialist dots today should expect a sales conversation, not a price list, for the foreseeable future.

How the competitors compare

Neither rival has published clearer numbers, which puts all three roughly at parity on the "how much will this actually cost me" question — genuinely unusual for three products launched within three weeks of each other. Grok Bot is bundled into existing SuperGrok and Cursor subscription tiers via a separate usage quota, with enterprise access still on a waitlist and no published per-bot pricing for business deployments. Muse is free for most features with paid tiers for additional usage — closer to a published structure than either competitor, though the specific thresholds and prices for those paid tiers aren't detailed in the sources reviewed either.

The pattern across all three, also visible in MacRumors' rundown of the dots launch: none of these companies is ready to commit to a public number for what happens once a user or business wants more agent than the free or bundled tier provides. That's likely not an accident. Pricing an entirely new product category — compute-metered, continuously running, unpredictable in actual usage per customer — before real usage data comes in is a genuinely hard problem, and every one of these three companies appears to be solving it the same way: ship first, price later, watch what people actually do with the free tier before committing to numbers that are hard to walk back once published.

For teams trying to plan a budget today, the practical takeaway is to treat the free dot, the free Grok Bot allotment, or Muse's free tier as an evaluation period rather than a stable cost baseline — and to expect the real pricing conversation, for all three products, to arrive sometime in the next two quarters once usage patterns settle. It's the same reason aiskill.market's own approach to skills pricing stays simple: a capability you install once, inspect, and run on infrastructure you already pay for doesn't carry the same open pricing question a standing cloud agent does.

Part 27 of 30 in the ChatGPT Dots series. Previously: Will Dots Replace Claude Code Skills, or Just Sit Beside Them?. Next: The Activity View: How Much Oversight Is Actually Practical?. Browse more builder insights or explore AI skills at aiskill.market.