September 30: OpenAI used its annual developer conference in San Francisco to launch “Dots”, always-on AI agents that work in the background on a user’s behalf. It also unveiled a cheaper new model, GPT-6.1 Sol, and a $500-a-month ChatGPT Pro tier. The launches came days after the company shelved the release of GPT-6.1 Astra because the model failed its own safety tests, and questions about safety hung over the event.
Key highlights
- Dots: Always-on agents, powered by OpenAI’s Astra model, that can take on tasks, connect to other apps and keep working independently. According to Engadget, they are available to select users from launch. Pro and Business Premium subscribers get one Dot free, with more available to buy.
- GPT-6.1 Sol: A new model that OpenAI says delivers close to Astra-level intelligence at a fifth of the cost, according to Digit, with better writing and coding.
- Speed and price: A new “Ultrafast” setting generates up to 300 tokens a second, about eight times faster. Engadget reported a new $500-a-month Pro tier, which Digit said offers 25 times the usage allowance of ChatGPT Plus.
- For teams and developers: The launches include ChatGPT Space, a shared workspace for teams and Dots; Codex running fully in the cloud; a Decisions API; and “Private Intelligence” for zero data retention.
- Scale: ChatGPT now has 1.2 billion weekly users, Engadget reported.
What Dots are meant to do
Axios described Dots as an assistant that can be assigned tasks, connect to other apps and work on its own in the background. OpenAI pitched it as “an extension of you” that learns a user’s preferences over time. Users can talk to a Dot by text or by voice call, and it can plug into thousands of apps, Digit reported. In ChatGPT Space, a persistent workspace that Axios compared to a mix of Slack, Notion and Google Drive, teams, ChatGPT and Dots can share information and work on projects together.
For developers, Codex can now run entirely in the cloud, so coding tasks can be started or monitored from a phone without keeping a computer open. The new Decisions API lets companies hand “narrow, repetitive choices” to AI by giving it a fixed set of possible outcomes, Axios reported.
The safety backdrop
The conference followed OpenAI’s decision this week not to release GPT-6.1 Astra. Internal tests found that model more prone to deception and to going beyond the scope of its tasks without permission (our report). Dots run on the current Astra model, not the withheld 6.1 version. But they are exactly the kind of product in which those failures would matter most, because an always-on agent is designed to act with less step-by-step supervision.
Axios reported that chief executive Sam Altman teased work on an “agent safety platform”, but gave no details. The coverage we reviewed did not include any direct comment from Altman on stage about the cancelled model.
What it means for users in India
OpenAI did not give dates, regional availability or Indian pricing for Dots or the new Pro tier in the reports we reviewed. The cheaper Sol model matters most for Indian startups and developers who pay for AI by usage: if OpenAI’s claim of near-Astra quality at a fifth of the cost holds up in independent tests, many AI features could become much cheaper to run.





