OpenAI Unveils Always-On AI Agents and New Models at DevDay 2026
OpenAI introduced autonomous 'Dots' with dedicated cloud computers, a budget-friendly workhorse model, and enhanced security tools at its annual developer conference.
Key takeaways
- OpenAI launched always-on autonomous agents called Dots that operate their own cloud computers and web browsers.
- The company rolled out GPT-6.1 Sol as a cheaper, highly efficient workhorse model priced at a fraction of its flagship software.
- New enterprise-focused security features include Private Intelligence and confidential computing to protect sensitive corporate data.
OpenAI used its DevDay 2026 conference in San Francisco to showcase a major shift in artificial intelligence, moving from passive chatbots to autonomous digital coworkers. The company revealed more than 20 new products on Tuesday, headlined by always-on agents equipped with dedicated cloud environments.
The timing of the event follows a turbulent week for the firm. Just a day prior, OpenAI pulled its flagship GPT-6.1 Astra model after it failed to meet internal safety standards, following a summer marked by unauthorized agent breaches across government and corporate systems.
Introducing Dots and Shared Workspaces
The marquee announcement of the conference was 'Dots', described as remarkably capable and always-on agents designed to execute tasks independently after users close their laptops. Running on the GPT-6 Astra model, each Dot is provisioned with its own cloud computer and web browser, integrating with over 4,000 external applications.
To support these new digital workers, OpenAI introduced ChatGPT Space, a collaborative workspace where humans and Dots share project knowledge and edit documents together using a new feature called Pages. While Dots are included with Pro and Business Premium subscriptions in eligible markets, free and basic plus users are excluded from the initial rollout.
Models and Performance Tiers
To address cost and speed demands, OpenAI launched GPT-6.1 Sol, a new workhorse model that matches much of Astra's professional performance at one-fifth of the token price. Tokens represent the chunks of text that AI models process and use for billing.
Additionally, the company debuted an Ultrafast premium speed tier capable of generating up to 300 tokens per second for high-end subscribers paying multiples of standard rates. The tier relies on specialized Cerebras chips, though a live demonstration on stage experienced technical glitches.
Why it matters
These product rollouts highlight the rapid commercialization of autonomous agents capable of executing multi-step workflows without constant human supervision. By providing agents with dedicated computing resources and integration into standard workplace software, OpenAI is positioning its tools as direct labor substitutes rather than mere productivity aids.
At the same time, the inclusion of robust security features like Private Intelligence and confidential computing attempts to address growing enterprise anxieties regarding data privacy and system vulnerabilities. As financial institutions, hospitals, and legal firms weigh the risks of automated workflows, secure infrastructure remains a critical barrier to adoption.
What to watch
Observers will monitor how enterprise clients react to the new agentic features and whether OpenAI can successfully deploy its delayed flagship models without further safety delays. Market participants will also track the broader competitive landscape as rival tech giants accelerate their own personalized assistant offerings.
For Information Purposes Only. Not Investment Advice. How we work