OpenAI used its DevDay 2026 event to introduce a new class of always-on AI agents called "dots," alongside a cheaper version of its flagship model and a $500 speed tier. The announcements, made during the company's annual developer conference, signal a push toward persistent, autonomous computing that runs continuously rather than waiting for a prompt.
The name "dots" refers to agents that come with their own computers. That detail matters more than the branding: it means each dot operates independently, with its own environment, rather than borrowing a slice of a shared session. OpenAI didn't say how many dots a single user could run or what hardware backs them.
What 'dots' actually are
Always-on agents have been a talking point across the AI industry for a while, but shipping them is a different problem. A dot that has its own computer can keep working after a user closes a laptop or moves to another task. In practice, that could mean an agent that monitors an inbox, runs scheduled jobs, or maintains a long-running process without a human restarting it every few hours.
OpenAI hasn't published a detailed spec sheet for dots, so questions about isolation, resource limits, and pricing remain open. The company framed dots as agents with their own computers, not as a shared pool. That's the whole pitch: persistence as a product feature.
Developers in the room will want to know how dots are billed and how long they can run unattended. Those answers weren't part of the announcement.
A cheaper GPT-6.1 Sol
The company also announced a cheaper GPT-6.1 Sol model. The existing Sol line has been positioned as a high-capability option, and a lower-cost variant would let developers put it into production workflows where per-call costs add up fast. OpenAI didn't disclose the new price per token or how the cheaper version differs from the original in capability.
Cheaper models tend to trade off some reasoning depth for cost. Whether the cheaper Sol makes that trade, and by how much, is the kind of detail that usually shows up in documentation after the keynote. For now, the headline is simple: a lower-cost entry point into the Sol family.
The $500 speed tier
OpenAI also introduced a $500 speed tier. The company didn't explain what the tier includes, what speed it delivers, or whether the fee is monthly or one-time. The figure itself is the announcement at this stage.
A premium speed tier suggests OpenAI sees demand from customers who care more about latency than price. That's a different buyer than the one the cheaper Sol model targets, and putting both on the same DevDay stage is a deliberate split: serve cost-sensitive developers on one side, latency-sensitive ones on the other.
Without a published rate card, the $500 figure is hard to evaluate against competing offerings. Developers will have to wait for specifics.
What's still missing
DevDay 2026 produced three product headlines and very few operational details. No launch dates were given for dots. No pricing breakdown was shared for the cheaper GPT-6.1 Sol. The $500 speed tier has a number attached to it but no feature list.
That's normal for a developer conference — keynotes set direction, documentation fills in the rest. But it does mean the practical questions are unanswered. How many dots can one account run? Does a dot keep running if the user's subscription lapses? Is the cheaper Sol available through the same API as the full model? Those are the things builders need before they can plan anything.
OpenAI has a pattern of following DevDay announcements with API documentation in the days and weeks after. The next concrete step is that documentation, and until it lands, dots, the cheaper Sol, and the $500 tier are names without terms.
