OpenAI Dots: Pricing Tiers, Access, and Usage Limits Explained
OpenAI's Dots agent needs ChatGPT Pro starting at $100/month. Here's how the $100, $200, and $500 tiers differ on usage and access.
What does OpenAI Dots cost?
Dots, OpenAI’s new persistent agent announced at DevDay, is gated behind paid ChatGPT plans. Your first Dot comes with eligible Pro plans starting at $100 a month. Moving up to $200 buys more usage headroom, and the $500 tier opens up the largest usage allowance along with access to Astra Ultra, OpenAI’s fastest model tier. There’s no free version of Dots, which separates it from competing personal-agent products that offer a no-cost entry point.
TL;DR
- Dots requires a paid plan, with the entry point set at $100/month for a ChatGPT Pro tier that includes your first Dot.
- Higher tiers buy usage, not just features: $200/month increases your usage allowance, and $500/month gives you the largest allowance plus access to Astra Ultra.
- You don’t need the top tier to get value, since GPT 6.1 Sol, a cheaper and more efficient model, handles a lot of serious work while barely touching weekly plan limits.
- Sol 6.1 is the quiet story of this release, reportedly approaching Astra’s performance on selected evaluations at a fraction of Astra’s standard token price.
- A Dot runs on top of ChatGPT, with its own cloud computer and memory, reachable from the web, mobile, Slack, and Codex.
- Context is opt-in, not automatic: Dots pulls from your ChatGPT memories and whatever connected apps you allow, rather than absorbing everything by default.
- Existing $200 subscribers face a plan change tied to this release that some users have pushed back on.
Other agents start typing. Remy starts asking.
Scoping, trade-offs, edge cases — the real work. Before a line of code.
How much does each Dots pricing tier actually get you?
OpenAI structured Dots access around three price points: $100, $200, and $500 a month. The $100 tier is the floor. It’s bundled into eligible ChatGPT Pro plans and includes your first Dot, meaning you don’t need to buy anything extra to start experimenting with an always-on agent.
The $200 tier exists primarily to buy more usage capacity. If you’re running an agent that works in the background, checks your calendar, cleans up recurring meetings, or drafts revisions to documents while you’re away from your desk, you’ll burn through usage allowances faster than with a normal chat session. The jump to $200 is aimed at people who hit those limits regularly.
The $500 tier is the ceiling. It comes with the largest usage allowance and adds access to Astra Ultra, OpenAI’s highest-speed model configuration. Astra Ultra is built for throughput. Think of it as the option for users who need an agent (or several) executing complex, multi-step work continuously rather than answering occasional queries.
None of this means you need the $500 plan to get real value out of a Dot. A lot of the day-to-day utility, catching a zombie recurring meeting, flagging a scheduling conflict, prepping revisions when a product launch changes scope, doesn’t require the top-end model. It requires an agent that has context and checks in regularly.
Is the $500 Astra Ultra tier worth it?
For most users, probably not, at least not yet. The usage math changes significantly once you factor in GPT 6.1 Sol, the efficiency-focused model OpenAI released alongside Dots. Sol is positioned as a cheaper alternative to Astra that still handles demanding work: complicated documents, code tasks, and research. In practice, running Sol for most of your agent work leaves most of your weekly plan allowance untouched, often dropping by just one or two percentage points per task.
OpenAI has reported that Sol approaches Astra’s performance on selected evaluations while costing a fifth of Astra’s standard input and output token prices. That pricing gap is exactly why it chews through so little of your plan allowance. If your Dot is mostly doing calendar hygiene, document drafts, and routine checks, Sol can likely carry that load. Astra and Astra Ultra make more sense when a task genuinely needs frontier-level reasoning, which is a smaller slice of daily agent work than the marketing around “agents with computers and memory” might suggest.
The practical advice: look at which model your Dot is actually using for the bulk of its tasks before assuming you need to pay for more capacity. Upgrading the model you route work to can matter more than upgrading the plan tier itself.
What do you actually get for the money?
One coffee. One working app.
You bring the idea. Remy manages the project.
A Dot is more than a chatbot with a longer memory. Each Dot runs on top of ChatGPT (on the Astra model line by default), with its own cloud-based computer and its own persistent memory. You can reach it from multiple surfaces: the ChatGPT web app, mobile, Slack, and Codex, so it’s not confined to a single chat window. You assign it ongoing responsibilities and connect it to the applications you want it to watch or act on, like a calendar, a messaging tool, or a document workspace.
That’s paired with two other DevDay announcements that shape the value proposition. Spaces, a shared workspace for people and AI to collaborate on documents (called pages) in a format that resembles collaborative office suites. And Codex gaining reusable cloud environments for software development, letting coding agents persist work without needing a laptop left open and running.
Together, these form OpenAI’s bet that more of a user’s actual work, documents, decisions, scheduling, code, should live inside ChatGPT over time, so the system accumulates more context and can act on it with less hand-holding.
How does Dots handle privacy and context boundaries?
Despite the “intelligent context” pitch, Dots doesn’t automatically ingest everything you’ve ever done in ChatGPT. It receives your existing ChatGPT memories and can build new memories from whatever applications you explicitly connect to it. That’s a deliberate boundary rather than a technical limitation, and it functions as an opt-in model: you decide what parts of your work context a Dot can see by choosing what to connect.
This matters for anyone wary of agents quietly absorbing every Slack message or calendar entry. The tradeoff is that a Dot is only as useful as what you let it access. If you want it to be a strong agent for work, connecting it to the tools you actually use for work, rather than leaving it limited to casual chat history, is what makes the difference.
How does Dots compare to other personal agents on the market?
Dots isn’t the only product combining a persistent computer, memory, and background execution. Meta’s Muse offers a similar setup and, notably, includes a free tier that Dots lacks. Muse has leaned into practical, lower-complexity tasks: email handling, small shopping tasks, and making phone calls on a user’s behalf, the kind of life-admin work people avoid doing themselves. Another entrant, Instinct, leans more heavily into travel and dining transactions, with reports that a notable share of users add a payment method to the platform within weeks of signing up.
The underlying form factor, a computer plus memory plus background work, is converging across these products. What isn’t converging, at least not yet, is utility. Each of these agents is being tuned toward different jobs: Muse toward everyday personal admin, Instinct toward travel and transactions, Dots toward professional context and recurring work tasks. The pricing structure reflects that focus. Dots has no free tier because OpenAI is positioning it as a work tool tied to paid ChatGPT plans, not a mass-market convenience app.
Frequently Asked Questions
How much does OpenAI Dots cost?
Dots is included with eligible ChatGPT Pro plans starting at $100 a month for your first Dot. Higher tiers, $200 and $500 a month, increase usage allowances, with the $500 tier also opening up access to Astra Ultra.
Is there a free version of Dots?
No. Unlike some competing agent products, OpenAI has not offered a free tier for Dots. Access starts at the $100/month Pro plan level.
Do I need the $500 plan to use Dots effectively?
Not necessarily. Much of the practical value, calendar fixes, document prep, scheduling conflict checks, runs well on GPT 6.1 Sol, a cheaper model that uses a small fraction of your plan’s usage allowance compared to Astra.
What’s the difference between Astra and GPT 6.1 Sol for Dots usage?
Astra is OpenAI’s stronger, more expensive model, used by default for Dots. GPT 6.1 Sol is a more efficient model that reportedly approaches Astra’s performance on selected evaluations while costing a fifth as much per token, making it far lighter on usage limits.
Does Dots have access to everything in my ChatGPT history?
No. A Dot receives your existing ChatGPT memories and builds additional memory only from applications you explicitly connect, such as Slack or a calendar. It does not automatically pull in all prior ChatGPT activity.
