At DevDay on September 29, OpenAI announced more than twenty updates, but two of them matter most to anyone building automations for clients: Dots, always-on agents that run on their own cloud computers, and GPT-6.1 Sol, a model that reportedly delivers near-Astra performance at roughly one-fifth of Astra's price. Together they change what "an AI teammate" can look like in a real workflow, and how much it costs to run one.
Dots are persistent agents powered by GPT-6 Astra. Each one gets its own cloud computer and browser, keeps continuity across projects, and can connect to more than 4,000 apps. You can talk to a dot by message or call on desktop, web and mobile, and through Slack and Microsoft Teams (texting is listed as coming soon).
The interesting part is the proactive behaviour. When you are away, a dot can do "proactive research" across the apps you have connected. One early tester told OpenAI his dot noticed an invoice he needed to send, pulled the details from an email thread and drafted it. OpenAI also pitches specialist dots with their own credentials and tools, for example one that watches customer feedback and another that reruns an analysis when new data lands.
OpenAI was careful about limits, and that is what agencies should study:
If you sell agent-based automation, this is a useful template for your own client deployments: read-only by default, explicit approval for anything that writes, and a rule set the client can inspect.
Dots are rolling out to ChatGPT Pro and Business Premium in eligible markets, with Pro users in the EEA, Switzerland and the UK currently excluded. Enterprise, Edu and Healthcare workspaces get a beta once an admin enables it. The first dot is included at no extra charge, though the underlying work in Codex or ChatGPT Work still uses your normal allowances.
GPT-6.1 Sol is priced at $2 per million input tokens and $10 per million output tokens, with cached input dropping to $0.10 per million. That is about one-fifth of GPT-6 Astra's $10 / $50 pricing. OpenAI reports it matches Astra on DeepSWE v1.1 and lands within about two points on OSWorld 2.0, and that it beats Claude Opus 5.5 on AutomationBench by 2.2 points at lower cost. These are vendor-reported numbers, and Hacker News commenters were already debating benchmark reliability, so test on your own workloads before switching anything.
There is also an Ultrafast tier delivering up to 300 tokens per second at roughly six times standard pricing, powered by Cerebras chips. It is live for Astra, with a Sol version promised "in the coming days", and it sits behind the new $500/month Pro 500 plan and Enterprise.
TechCrunch framed the app and plugin changes as a direct play at the app-store model, with ChatGPT suggesting and running apps inside the conversation. No revenue-sharing mechanism has been announced yet.
Competition is tight. Anthropic shipped Claude Sonnet 5.5 the day before at an unchanged $2 / $10 price, and OpenAI separately shelved GPT-6.1 Astra over safety evaluation results. Frontier pricing is falling quickly, which is good news for margins on agent-heavy client work.
Our cutting-edge features simplify collaboration and creativity, making your workflow intuitive and efficient. Transform your vision into reality effortlessly with Hadidiz Flow.
