OpenAI Officially Releases GPT-6 Astra: Next-Generation Flagship with Critical Cyber Safeguards
On September 3, 2026, OpenAI officially announced the deployment of GPT-6 Astra, its next-generation flagship foundation model, following a weeks-long safety pause implemented after initial frontier evaluations revealed elevated autonomous capability.
GPT-6 Astra represents OpenAI's primary generational leap beyond the GPT-5.6 family, uniting multimodal vision, native tool use, and advanced mathematical reasoning into a unified architecture. According to OpenAI's published technical overview and system card, the model recorded a state-of-the-art 72.6% success rate on the OSWorld computer-use benchmark, demonstrating the ability to navigate desktop graphical user interfaces, execute complex multi-step office workflows, and operate third-party software directly without bespoke API integrations.
Addressing autonomous exploit concerns identified during pre-deployment vetting, OpenAI published its comprehensive "Path to Astra" safety framework alongside the release. The document details new multi-layer containment safeguards and real-time behavioral monitoring designed to mitigate autonomous cyber risks, satisfying threshold requirements under the voluntary frontier AI safety vetting framework finalized in August.
API pricing for GPT-6 Astra is set at $10.00 per million input tokens and $50.00 per million output tokens, aligning directly with Anthropic's Claude Fable 5.1. Rollout begins immediately as a limited preview for enterprise partners and safety researchers, with general availability across ChatGPT paid tiers (Plus, Pro, Enterprise) rolling out in stages over the coming days.
