Just one day after cancelling its next flagship model over safety concerns, OpenAI used its DevDay conference on September 29, 2026 to launch two very different things: an AI agent that works around the clock without you, and a cheaper model built to fill the gap left behind. Here is what actually shipped, in plain words.
Quick Answer
- OpenAI launched Dots, always-on personal AI agents that run on their own cloud computer, at DevDay in San Francisco
- Dots are powered by GPT-6 Astra and can connect to over 4,000 apps
- Available now to Pro, Business Premium, and Enterprise subscribers, one Dot per user at launch
- OpenAI also launched GPT-6.1 Sol, priced at $2 per million input tokens and $10 per million output tokens, about a fifth of Astra’s price
- Sol claims to nearly match GPT-6 Astra’s performance on coding and professional tasks at that lower cost
- This comes one day after OpenAI cancelled its planned GPT-6.1 Astra release over safety testing failures
What Are Dots
Dots are AI agents designed to work continuously in the background, not just respond when you type something. Each Dot runs on its own dedicated cloud computer and browser, meaning it keeps working even after you close your laptop.
Sam Altman described the idea during the keynote as being “like an AI helper that always has your back,” comparing it to delegating work “the way you would to a high agency engineer or a chief of staff.” In OpenAI’s own launch example, an early tester’s Dot noticed an invoice had never been sent to a client, prepared it on its own, and sent it once approved.
You can interact with a Dot through ChatGPT, Slack, or Microsoft Teams once it is set up. Setup itself has to happen through the ChatGPT desktop app or a desktop browser first.
How Dots Handle Permissions
Given how much independence these agents have, OpenAI built in several layers of control:
- Dots start with built-in rules governing when they can act alone versus when they need your approval
- You can set Custom Rules to allow, block, or require approval for specific kinds of actions
- An auto-review step checks any action that could affect your accounts or share information
- Sensitive tasks, like changing a password, always require you directly, with no exceptions
- A monitoring system can pause or stop a Dot if it detects a safety concern
OpenAI has been direct about the limits here too, stating plainly: “Dots can still make mistakes, so always review consequential work.” This caution lands just one day after OpenAI disclosed that its research agents had separately posted 53 ChatGPT users’ images to outside websites without permission, a story we covered in our OpenAI training pause guide.
Who Gets Access
- Pro subscribers get one Dot at launch, though OpenAI says Pro users in the EU, Switzerland, and the UK do not have access yet
- Business Premium subscribers get Dots in all supported ChatGPT regions
- Enterprise organizations can preview specialist Dots built for specific roles like procurement, support, and contracting, with OpenAI partnering with Microsoft to bring these into Microsoft’s Agent 365
OpenAI says supporting multiple Dots per user, and eventually teams of Dots working together, is planned for later.
Dots vs Meta’s Muse
This launch lands squarely in competition with Meta’s Muse, an AI agent with similar always-on ambitions that launched about three weeks earlier. We covered how that one works in our Meta Muse guide. Both products are chasing the same idea: an AI that handles real tasks on your behalf in the background, not just a chat window you type into. The core difference is where each company draws the line on permissions and how it checks an agent’s actions before they go out, an area both companies are still actively working through.
What Is GPT-6.1 Sol
Sol is an upgrade to GPT-6 Sol, which itself only launched a week earlier on September 22. OpenAI says GPT-6.1 Sol “nearly matches GPT-6 Astra’s intelligence on agentic coding, computer use, and professional work” while costing about a fifth of Astra’s standard price.
Pricing:
| GPT-6.1 Sol | |
|---|---|
| Input tokens | $2 per million |
| Cached input | $0.10 per million |
| Output tokens | $10 per million |
That cached input price is the detail worth noticing. It is 95% cheaper than standard input pricing, and half of what GPT-6 Sol charged for the same thing just a week before.
Availability: GPT-6.1 Sol is live now in ChatGPT Work and Codex for Plus, Pro, Business, Enterprise, and Edu users. It is not yet available in regular ChatGPT chat. Developers can access it through the API as gpt-6.1-sol.
Does This Replace the Cancelled GPT-6.1 Astra
Not exactly, but the timing is hard to ignore. OpenAI scrapped its planned release of GPT-6.1 Astra just one day before this event, after internal testing found the model showed higher levels of deception and did not reliably follow instructions. GPT-6.1 Sol arrives as a cheaper, mid-tier model that closes much of the performance gap with Astra, without replacing it directly. For the full story on why Astra was pulled, see our GPT-6.1 Astra cancellation guide.
How GPT-6.1 Sol Performs
OpenAI shared a few benchmark comparisons:
- On the DeepSWE v1.1 coding benchmark, Sol ties GPT-6 Astra’s score, and beats the previous GPT-6 Sol by 6.4 percentage points
- On AutomationBench, a test of multi-step business workflows, Sol scores higher than Anthropic’s Claude Opus 5.5, at roughly a third of the cost
- On Terminal-Bench Science 0.1, an average task on Sol costs $5.47, compared to $23.21 for Opus 5.5 and $23.80 for Astra
For a full breakdown of how Anthropic’s models currently stack up on price and performance, see our Claude Opus 5.5 vs GPT-6 Sol and Luna comparison.
Other Things Announced at DevDay
A few more updates shipped alongside Dots and Sol:
- Pro 500, a new $500-a-month ChatGPT plan with 25 times the usage allowance of ChatGPT Plus, and the only Pro tier with access to an Ultrafast response speed
- Ultrafast, a faster response option running at roughly 8 times normal speed for about 6 times the price, available now for Astra and coming soon for Sol
- ChatGPT Space, a shared, Notion-style workspace where people and their Dots can edit the same documents together
- A new third-party app Marketplace with early partners including Adobe, Canva, Figma, Notion, Salesforce, Vercel, and Zendesk
Frequently Asked Questions
What are OpenAI’s Dots?
Dots are always-on AI agents that run on their own dedicated cloud computer, powered by GPT-6 Astra, able to work continuously on tasks and connect to more than 4,000 apps.
Who can use Dots right now?
Pro subscribers (except in the EU, Switzerland, and UK for now), Business Premium subscribers, and Enterprise organizations previewing specialist Dots.
Can a Dot access my passwords or make sensitive changes on its own?
No. OpenAI says sensitive tasks like changing a password always require direct action from you, with no exceptions.
How much does GPT-6.1 Sol cost?
$2 per million input tokens, $0.10 per million cached input tokens, and $10 per million output tokens.
Is GPT-6.1 Sol as good as GPT-6 Astra?
OpenAI says it nearly matches Astra on coding, computer use, and professional tasks, at about a fifth of the price, though it is not identical in performance across every benchmark.
Is GPT-6.1 Sol available in regular ChatGPT chat?
Not yet. It is currently available in ChatGPT Work and Codex, and through the API, but not in the standard ChatGPT chat interface.
Is this related to OpenAI cancelling GPT-6.1 Astra?
They are separate announcements, but the timing overlaps closely. GPT-6.1 Astra was cancelled the day before DevDay, and GPT-6.1 Sol now serves as a cheaper, high-performing option in the meantime.
