OpenAI has unveiled a new always-on artificial intelligence assistant called Dots, positioning it as a comprehensive personal and professional agent with access to over 4,000 different tools and services.
Unlike conventional AI chatbots, Dots is designed to operate continuously, meaning it keeps working even when a user’s devices are powered off or run out of battery.
The launch was the headline announcement among roughly 20 product and service reveals made during OpenAI’s annual developer showcase, known as Dev Day, which the company uses to present the cumulative output of its year-round engineering efforts.
Dots draws comparisons to both Meta’s Muse and Elon Musk’s Grok Bot, but attempts to bridge the gap between the two by handling both professional tasks and personal life management within a single agent.
Each user is assigned their own virtual computer, enabling Dots to operate desktop environments much like a human would, accessing tools fluidly rather than relying on rigid, manual API connections from earlier AI agent generations.
OpenAI has stated that user data will not be used to train larger models, and the system includes back-end encryption along with secure password storage as part of its privacy architecture.
Dots runs on GPT-6 Astra, OpenAI’s most capable model, which the company says is critical because a more intelligent underlying model produces a more intuitive agent capable of anticipating user needs in advance.
Also announced at Dev Day was GPT-6.1 Sol, a new model described as delivering near GPT-6 Astra-level intelligence at one-fifth of the cost, a significant reduction that arrives just weeks after the previous flagship model was released.
OpenAI additionally introduced an Ultra-Fast mode capable of processing at 300 tokens per second, offering near-instantaneous responses, though access is priced at $500 per month as a separate tier above the existing $200 monthly Pro plan.
The speed upgrade was made possible in part through OpenAI’s earlier investment in Cerebras, a chip company that has since gone public and whose hardware was specifically designed to accelerate AI model inference at scale.
A collaborative product called ChatGPT Space was also announced, described as a multiplayer version of ChatGPT where users can invite both friends and AI agents into a shared workspace to collaborate on tasks in real time.
OpenAI also demonstrated integration with Hugging Face’s Micro Duck robot, connecting GPT-6 and its Codex coding agent to the physical device, illustrating the company’s ambition to extend its software platform into real-world hardware and smart devices.
The overarching theme of Dev Day, according to commentary from In the Loop host Ejaaz Ahamadeen, is that OpenAI is not building isolated products but rather constructing an interconnected ecosystem that mirrors the kind of omnipresent personal AI famously depicted as Jarvis in the Iron Man franchise.
OpenAI’s multi-modal investment strategy, spanning agents, models, chips, robotics, and collaborative tools, is seen as its primary competitive advantage over rivals such as Anthropic, Meta, and SpaceX AI, each of which operates within narrower product verticals.
