OpenAI Launches GPT-6 Astra with Advanced 'Computer Use' Capabilities
The new flagship model shifts AI from a chatbot to a functional agent capable of operating desktop apps and websites.
OpenAI began the gradual rollout of its new flagship model, GPT-6 Astra, on September 3, 2026. The release marks a strategic pivot toward agentic AI, designed specifically to handle complex business workflows by interacting directly with computer interfaces.
Unlike previous iterations, GPT-6 Astra features "computer use" capabilities, allowing the model to operate across various websites and desktop applications without the need for dedicated APIs. This allows the AI to execute tasks within software such as Blender or internal corporate tools by navigating the UI like a human user. To support these expansive tasks, the model includes a massive context window of approximately 1.05 million tokens.
A Tiered Access Strategy
OpenAI has implemented a tiered rollout based on subscription plans, utilizing different branding and interfaces depending on the account type. For users on Pro, Business, and Enterprise plans, the model appears in the standard ChatGPT interface branded as "GPT-6 Pro." In contrast, Plus subscribers do not access the model through the regular chat interface; instead, they must use "ChatGPT Work" and "Codex" to utilize Astra's capabilities.
The Shift to Agentic AI
GPT-6 Astra succeeds GPT-5.6 Sol and represents what OpenAI describes as a generational leap in intelligence and alignment. By focusing on the ability to manipulate a computer's user interface, OpenAI is moving the technology beyond the constraints of a conversational chatbot. This transition allows the AI to function as an autonomous agent capable of executing end-to-end professional workflows across multiple disparate applications.
Industry Implications
This development significantly reduces the industry's reliance on the development of custom APIs for every software tool. By enabling an AI to "see" and interact with any software on a screen, OpenAI expands the scope of automatable professional tasks. This shift potentially redefines how humans interact with software, moving from manual input to high-level orchestration of AI agents.
Enterprise Validation
Early feedback from enterprise partners suggests the model is highly effective for professional environments. Yashodha Bhavnani, VP of AI Products at Box, noted that GPT-6 delivered some of the best and highest-signal recommendations the company has seen. As the rollout continues, the industry will be watching to see how these agentic capabilities scale across diverse operating systems and the security implications of allowing an AI to operate a desktop environment independently.