OpenAI rolls out GPT-6 Astra to Pro, Enterprise, and Business Premium users in ChatGPT Work and Codex and in the API

OpenAI has released GPT-6 Astra, a significant upgrade to its ChatGPT and Codex offerings, now available to Pro, Enterprise, and Business Premium users, as well as through the API and on AWS. GPT-6 Astra is described as the "world's most intelligent and aligned model," demonstrating state-of-the-art performance across computer use, software engineering, cybersecurity, and scientific research, even solving long-standing mathematical problems and achieving perfect scores on benchmarks like ExploitBench. A key advancement is Astra's enhanced computer use capability, allowing ChatGPT to operate complex workflows, generate polished documents, and execute tasks in the background on user desktops with nearly double the speed of previous models. For Codex, Astra introduces a novel context preservation method that avoids the lossy summarization of previous models, enabling better debugging and complex problem-solving by retaining searchable historical context across expanded context windows. This release follows a rapid iteration cycle for GPT models, building on the capabilities introduced with GPT-5.

AI Signal Decode

GPT-6 Astra represents a substantial leap in AI capabilities, particularly in "computer use," which now enables ChatGPT to manage intricate professional workflows, produce professional-grade documents, and operate at significantly increased speeds—nearly twice as fast as previous iterations. This enhanced functionality directly benefits users in professional fields, allowing for more complex tasks to be delegated to the AI. The model's performance metrics, including saturating FrontierMath Tier 4 and achieving near-perfect scores on ARC-AGI-3 and ExploitBench, underscore its advanced reasoning and problem-solving prowess, positioning it as a benchmark for future AI development.

The implications for the market are considerable, with OpenAI targeting professional and enterprise users who require advanced AI assistance for demanding tasks like software engineering and cybersecurity. The introduction of a new context preservation mechanism in Codex, which allows for searchable historical data across context windows rather than relying on potentially lossy summaries, is a critical development for long-form projects and complex debugging scenarios. This feature addresses a long-standing limitation in LLMs and could significantly boost productivity for developers and researchers.

Technically, GPT-6 Astra's advancements in 'computer use' suggest a deeper integration with operating system functionalities and a more robust ability to chain actions and applications. The improved context handling in Codex is a significant architectural improvement, moving beyond simple summarization to a more persistent and accessible memory. The slow rollout to specific user tiers, starting with premium business and enterprise accounts, indicates OpenAI's strategy to manage adoption and gather feedback for such a powerful and potentially disruptive technology, especially given its "significant jump in cyber capabilities."