OpenAI has introduced GPT‑6 Astra, a model designed for enterprise and professional workloads. Astra is now available in ChatGPT Work, Codex, and via the API. According to OpenAI, the model is particularly capable in computer use, browsing, professional tasks, software engineering, cybersecurity, and scientific work.
Integration with existing workflows
OpenAI says Astra can operate within the same daily applications people use—even when those applications lack an API—reducing the need for data preparation, workflow redesign, or custom integrations. Early customer use cases reported within days of rollout include GPU optimization, detecting discrepancies in financial statements, and producing on‑brand slide decks.
OpenAI deployed Astra internally weeks before the public launch. The company’s developer and marketing teams used Astra and Codex to turn three hours of multicamera footage into a “GPT‑6 Astra Developer First Impressions” video, which accumulated over 550,000 views in four days. Engineers used Astra to identify and fix a memory‑allocation bottleneck in a test environment; by switching allocators they achieved about 25× lower turn latency while peak memory usage rose roughly 30%.
Efficiency and pricing
Astra was trained to complete tasks using fewer tokens and with fewer retries, which OpenAI says reduces rework and lowers cost per task. The company reports that Astra occupies the majority of the cost‑efficiency frontier on professional work and coding benchmarks, citing Terminal Bench 4.0 and the Artificial Analysis Intelligence Index. Pricing begins at $10 per million input tokens and $50 per million output tokens.
Alignment, safety testing and administrative controls
OpenAI emphasizes Astra’s alignment with human intent and authorization. During training, Astra was evaluated on OpenAI’s internal computer use safety benchmark, which examines difficult business scenarios such as exposing confidential information, over‑sharing dashboards, or deleting data. In that evaluation Astra produced unintended outcomes 89% less often than GPT‑5.6 Sol and 74.7% less often than Claude Fable 5.1. Additional confirmation steps and automated review further improved performance for GPT‑6 Astra and GPT‑5.6 Sol.
New enterprise admin controls let organizations restrict Astra’s access to approved websites and desktop applications, manage uploads and downloads, and control browsing history. ChatGPT Work and Codex include safeguards such as confirmation policies that can require approval before consequential actions and automated review of potentially unsafe or unauthorized tool calls. These options are intended to allow organizations to start with limited configurations and expand access over time.
Cybersecurity and preparedness
Astra is the first model to reach the “Critical” cybersecurity capability threshold under OpenAI’s Preparedness Framework. To match that capability, OpenAI has strengthened protections against misuse and unauthorized actions by training Astra to respect safety and security boundaries, improving its resistance to guardrail bypass attempts, and deploying automated checks to block harmful responses.
Zero Data Retention is available for eligible API customers on supported endpoints, subject to approval.
Plugins and availability
Alongside Astra’s release, OpenAI launched enterprise plugins for ChatGPT Desktop from Oracle Analytics, Power BI (a Microsoft Fabric service), Navan, and Avalara to facilitate access to common enterprise applications using the model’s browser capabilities.
Astra can be tried in ChatGPT Work or Codex, or integrated into products and workflows via the API. Enterprise access is off by default at launch; administrators can enable Astra under the applicable rate card and agreement.
