OpenAI announced over 20 launches at DevDay 2026 in San Francisco, centering on "Dots," always-on AI agents powered by the flagship GPT-6 Astra model. These agents operate on dedicated cloud computers with web browsers, integrating with more than 4,000 applications to perform tasks autonomously while users are offline. The initial rollout is restricted to Pro and Business Premium subscribers in eligible markets, excluding Free and Plus tiers. Alongside Dots, OpenAI released ChatGPT Space, a collaborative workspace where human teams and their respective agents edit shared documents and manage project knowledge simultaneously.
The company also unveiled GPT-6.1 Sol, a new workhorse model that reportedly matches GPT-6 Astra’s performance in coding and computer use at one-fifth of the token price. This release follows the recent cancellation of the GPT-6.1 Astra launch due to safety concerns. To address latency issues inherent in agentic workflows, OpenAI introduced an Ultrafast tier, available on the $500-a-month Pro 500 plan, which increases generation speed up to eight times in Codex using Cerebras chips. Additional features include the Decisions API for rapid classification, open-source access to the Codex harness, and Private Intelligence tools offering zero data retention for regulated industries.
The introduction of Dots signifies a structural shift from conversational interfaces to autonomous operational units within enterprise software ecosystems. By assigning agents their own cloud infrastructure and persistent memory, OpenAI moves beyond simple tool-calling into sustained background execution. This architecture creates significant integration dependencies, as these agents require deep access to internal application stacks and communication platforms like Slack and Teams. The restriction of this capability to higher-tier subscriptions suggests a strategy to monetize high-value workflow automation rather than general consumer interaction, potentially widening the gap between standard user capabilities and enterprise-grade operational efficiency.
Simultaneously, the release of GPT-6.1 Sol at a fraction of Astra’s cost addresses the economic viability of scaling agent-based systems. High-frequency agentic tasks consume substantial tokens, making cost-per-action a critical metric for adoption. By decoupling performance parity from premium pricing, OpenAI lowers the barrier for developers to embed complex reasoning loops into their products. However, the concurrent emphasis on security measures, such as the open-sourcing of the Codex harness and the launch of Codex Security Cloud, indicates an acknowledgment of the risks associated with autonomous code execution. The industry must now balance the utility of self-directed agents against the expanded attack surface they introduce into corporate networks.
