OpenAI introduced permanent AI agents Dots and new model GPT-6.1 Sol

OpenAI presented a new system of personal AI agents Dots, capable of constantly working on user tasks, as well as a new model GPT-6.1 Sol. The company made the announcements on September 29 at the DevDay 2026 conference.

The main new product was Dots — permanent AI agents that can independently perform long-term tasks and monitor their development without needing a separate command for each step. In OpenAI, they are described as 'always-on' assistants working on behalf of the user around the clock.

Each Dot gets its own cloud computer and browser. The agent can connect to user applications through the OpenAI plugin ecosystem — the company claims support for more than 4,000 services. Possible scenarios include tracking customer feedback, preparing fixes for software, updating presentations and documents, analyzing new data, or preparing materials for publication. 

You can communicate with the agent through ChatGPT, as well as Slack and Microsoft Teams. Dots support voice interaction, and the company plans to add messaging via regular messages later. The agent maintains context across different communication channels and over time takes into account user preferences and requirements for the result.

At the same time, OpenAI has provided restrictions for autonomous work of agents. When the user is not directly interacting with the Dot, the agent can independently explore connected sources, but for such background work, applications are switched to read-only mode. Sending messages, changing data, and other significant actions may require separate permission from the owner.

At launch, the user is provided with one primary Dot. The system is beginning to roll out to ChatGPT Pro and Business Premium subscribers on supported markets. Enterprise customers, as well as organizations on Edu and Healthcare plans, will be able to join the beta version after workspace administrator approval. In the future, OpenAI expects to allow users to create multiple interacting agents.

Dots run on the flagship GPT-6 Astra model. At the same time, OpenAI introduced a separate new model GPT-6.1 Sol — an updated version of GPT-6 Sol, focused primarily on programming, computer use, and executing multi-step professional tasks.

The company claims that on a number of tests GPT-6.1 Sol approached the capabilities of the more powerful GPT-6 Astra, but is significantly cheaper. In particular, the standard API cost is $2 per million input tokens, $0.10 per million cached input tokens, and $10 per million output tokens.

According to OpenAI's internal tests, the new Sol significantly improved results in programming, executing business processes, working with computer interfaces, and analyzing complex documents. The company also claims a reduction in factual errors and improved compliance with user constraints compared to the previous GPT-6 Sol. These metrics are based on OpenAI's own tests and may differ in real-world usage.

GPT-6.1 Sol is already becoming available to Plus, Pro, Business, Enterprise, and Edu users in ChatGPT Work and Codex, and developers can use it via the API. In regular Chat mode, the model is not yet available at launch. In the near future, OpenAI also promises a Sol Ultrafast version with token generation up to eight times faster than the standard mode in Codex.

In addition to Dots and GPT-6.1 Sol, at DevDay the company presented the ChatGPT Space team workspace, new Codex features, tools for creating custom plugins, and a number of enterprise solutions. In total, OpenAI announced more than 20 major product and platform updates.