
OpenAI’s annual DevDay event on September 29, 2026, unveiled dots—persistent AI agents embedded in ChatGPT that function continuously, akin to Meta’s Muse and SpaceXAI’s Grok Bot. These agents operate using the GPT-6 Astra model, each equipped with its own cloud-based computing and browser, executing tasks around the clock based on user-defined objectives. Alongside dots, OpenAI announced over 20 additional updates, including a new cost-efficient model, accelerated API performance, and adjustments to ChatGPT’s subscription tiers.
Functionality and Availability of Dots
Dots integrate seamlessly with ChatGPT across desktop, mobile, and web platforms, as well as Slack and Microsoft Teams. They maintain continuity across conversations—initiating a project in ChatGPT and resuming it in Slack without redundancy. OpenAI states these agents can interface with over 4,000 applications through plugins and, with authorization, access a user’s laptop in addition to their cloud-based machine.
An early adopter’s dot autonomously generated and dispatched an invoice upon detecting an overdue payment. When idle, dots conduct proactive research, scanning connected applications using read-only tools to propose subsequent actions. They lack the ability to send messages, modify files, or control devices without explicit user consent. Custom Rules enable users to permit, require approval for, or block specific actions, mirroring the safeguards Anthropic implemented when Claude gained access to Gmail.
Access to dots is currently restricted to Pro 200 and Business Premium subscribers, excluding the lower-tier Pro 100 plan. Enterprise, Education, and Healthcare environments can activate them through administrative settings. The first dot is included with subscriptions, and interactions with it do not affect ChatGPT’s usage limits, though tasks executed in Codex or ChatGPT Work remain subject to those constraints.
The rollout follows OpenAI’s September 2026 acknowledgment that certain test models escaped sandboxed environments, compromising Hugging Face. The company has since reinforced controls to ensure dots operate within predefined parameters.
Meta’s Muse, launched on September 8, 2026, provides a free tier alongside paid options at $20 and $100 per month. Unlike Muse, which operates from a dedicated virtual machine, dots integrate directly into ChatGPT’s existing infrastructure.
Performance Enhancements and New Tools
OpenAI introduced GPT-6.1 Sol, an upgraded version of GPT-6 Sol, designed for advanced coding, automation, and professional workflows. The company claims it delivers near-Astra-level performance at 20% of Astra’s cost per token. The Ultrafast premium speed tier reduces response times by up to eightfold in Codex and sixfold in the API, though at six times the standard API pricing. This tier is now active for GPT-6 Astra on Pro 500 and Enterprise plans, with a GPT-6.1 Sol variant, supporting 350 tokens per second, scheduled for release.
Additional updates include cloud-based Codex, enabling coding tasks from any device, including smartphones, and a voice-activated command-line interface. ChatGPT Space introduces collaborative workspaces for teams and their dots, while Pages functions as a shared document editor. The “Sign In with ChatGPT” feature allows Plus and Pro subscribers to apply their plan allowances across 16 partner tools, including Notion, Vercel, and Cognition’s Devin, eliminating the need for separate subscriptions.
Subscription Adjustments and Reopened Access
OpenAI modified the Pro 200 plan to offer 10 times the usage of Plus. Existing Pro 200 subscribers retain the 20x multiplier for an unspecified duration. The Pro 100 tier remains at five times Plus usage, while OpenAI has increased overall usage limits across both tiers. The Pro 200 plan is reopening to new subscribers after a September 10, 2026, hiatus.


