OpenAI’s GPT-6 Astra Brings Computer-Using AI Into Reach for Professional Work

OpenAI has introduced GPT-6 Astra, a flagship model designed to handle computer tasks, software engineering, research and other professional workflows. The model is rolling out through ChatGPT Work, Codex and the API, with enterprise customers receiving early access before broader availability across paid ChatGPT plans and Amazon Web Services.

Astra’s central pitch is practical execution. It can browse websites, operate computer interfaces, create documents, assist with coding and complete multi-step workflows that previously required a person to move between applications. OpenAI describes the model as its most capable and aligned system for business use, while Fox Business highlighted the model’s initial enterprise rollout and expansion timetable.

GPT-6 Astra’s defining capability is computer use, allowing it to interact with software and websites rather than only generating text in a chat window. OpenAI says the model sets state-of-the-art results on Agents’ Last Exam, AutomationBench and ScreenSpot Pro, benchmarks focused on computer workflows, browsing and visual interface understanding.

That capability shifts the unit of AI assistance from an answer to a completed task. A user can ask Astra to research information, draft a document, update software or navigate a browser-based process, with the model handling more of the intermediate steps. Community OpenAI materials also emphasize improved understanding of user intent, a critical safeguard when an automated system acts across multiple applications.

Astra targets work that combines reasoning, judgment and execution. OpenAI lists software engineering, cybersecurity, science, writing, design and professional analysis among the areas where the model is intended to perform at a high level.

The business case rests on reducing the setup required for useful AI. OpenAI’s work-focused announcement says companies can take on demanding projects without preparing every data source, redesigning every workflow or building a custom integration for each task. That promise places Astra closer to an AI operator than a conventional text assistant, although its value will depend on accuracy, permissions and the quality of the systems it can access. Eesel’s analysis frames the same change around cost and access, while VentureBeat presents the launch as part of OpenAI’s broader push toward agentic work.

GPT-6 Astra’s distribution strategy starts with organizations and expands toward individual users and developers. Enterprise customers with Daybreak access received the initial rollout, followed by ChatGPT Plus, Pro, Business and Enterprise users, along with API access and availability through Amazon Web Services.

The model’s reach will depend on more than technical performance. Businesses need predictable costs, secure permissions and clear controls before allowing an AI system to operate inside work applications. Eesel describes reported pricing at $10 and $50 tiers, while OpenAI’s announcements emphasize broad access across consumer, business and developer products. That combination puts pressure on rivals to offer comparable computer-use capabilities without making deployment too expensive or risky.

OpenAI and its supporters have framed Astra as a step into an AGI era, but the concrete evidence in the launch centers on benchmarks and workflow performance. VentureBeat’s launch coverage used the AGI framing, and Eesel reported that OpenAI president Greg Brockman said he personally believed the system could qualify as AGI.

Those claims carry weight because Astra combines several capabilities in one model, including browsing, coding, scientific work and computer control. They also require careful interpretation. Benchmark leadership does not by itself establish that a system can perform every form of human work reliably, and automation creates new failure modes when an agent takes actions rather than merely offering suggestions. The practical test will be sustained performance in real organizations, under supervision and across messy software environments.

Developers now have a model aimed at building agents that can see screens, navigate browsers, write software and complete professional tasks through the API. OpenAI’s simultaneous release across ChatGPT, Codex and API products gives builders several paths to test the same core capabilities, from individual workflows to enterprise applications.

The next phase will focus on reliability, access controls and operating cost. Developers will need to decide which actions an agent can take independently, where human approval is required and how failures are logged. OpenAI’s alignment claims and Astra’s emphasis on user intent address part of that problem, but adoption will depend on observed behavior in production. If the system performs consistently, more work that was previously blocked by time, staffing or specialized expertise becomes economically viable.

[sources:OpenAI Blog,Foxbusiness,Community.openai,Eesel,VentureBeat AI]