The announcements at OpenAI’s DevDay in San Francisco on September 29 centred on Dots, persistent agents powered by GPT-6 Astra. Unlike a conventional chatbot session, a Dot can continue working between conversations, use its own cloud computer and browser, draw on approved connected applications and return to the user when a decision or approval is required.
OpenAI said users can give a Dot an ongoing goal, connect the applications it needs and specify what it may do independently. The company is gradually rolling out the feature to eligible accounts. Its documentation says the agents can maintain progress across continuing assignments rather than waiting for a fresh prompt each time.
That autonomy also raises the stakes for permissions and security. OpenAI says users can define action boundaries and review consequential steps, while business data is not used to train its models without consent. The safeguards are significant because persistent agents may handle information and actions across email, calendars, workplace software and other connected services.
Alongside Dots, OpenAI introduced GPT-6.1 Sol, an upgraded reasoning model aimed at coding, computer use and professional work. The company says it approaches the performance of its flagship GPT-6 Astra on those tasks while charging one-fifth of Astra’s standard API input and output token prices.
GPT-6.1 Sol is priced at $2 per million input tokens and $10 per million output tokens, with cached input costing $0.10 per million tokens. OpenAI says that cached rate is 50 per cent below GPT-6 Sol’s and is designed to make context-heavy agent workloads cheaper to operate.
The model has a 1.05-million-token context window and supports up to 128,000 output tokens, according to OpenAI’s developer documentation. It is available through the API and is being rolled out in ChatGPT Work and Codex, beginning with Pro users and extending to Plus, Business, Enterprise and Edu customers.
DevDay also highlighted OpenAI’s effort to make ChatGPT a distribution layer for software. The company now operates a universal plugin directory shared by ChatGPT and Codex, where developers can publish packages that combine reusable skills, connections to external systems through the Model Context Protocol and, where appropriate, interactive interfaces.
Public plugins can be published once and discovered from supported surfaces across both ChatGPT and Codex. OpenAI’s developer rules say higher-quality plugins may receive greater distribution through directory placement or proactive suggestions inside conversations. That gives developers a route to reach users at the moment a particular service or workflow becomes relevant, rather than requiring them to open a separate application.
The approach builds on OpenAI’s earlier Apps SDK initiative, which allowed services to appear conversationally inside ChatGPT. The current plugin architecture broadens that concept by packaging applications, tools and repeatable workflows under a single installable unit, with authentication and workspace permissions retained for services that require access to external accounts.
OpenAI also announced ChatGPT Space, a collaborative environment that brings files and Pages together so people and AI can work on shared material. Users can draft and revise content, edit work directly and invite others to view, comment or make changes. The company also expanded ChatGPT Work across devices, allowing tasks started on desktop to continue on web or mobile.
For developers, the company placed its Agents API into public beta with hosted execution, memory, tools and multi-agent support. Computer-use capabilities allow agents to operate software through graphical interfaces, while a Decisions API entered limited preview for classification, routing and selecting actions from predefined choices.
Follow Arabian Post
Select Arabian Post as your preferred source on Google and MSN News for trusted business news and Arab politics and updates.
