OpenAI’s GPT-6 Astra launched in early September 2026 with broad paid-user and API access, yet immediate safety and capacity constraints have shaped limited rollout. Internal tests flagged the model’s critical cybersecurity capabilities, including zero-day exploitation potential, prompting tighter safeguards, sandboxed testing, and pauses on new $200 Pro subscriptions to manage compute strain. The planned October GPT-6.1 Astra update was canceled after alignment failures, while enterprise controls and phased defensive-only access further restrict general availability. These dynamics, alongside ongoing evaluations and competitive pressure from models like Gemini 4, sustain trader consensus that full public removal remains unlikely by late October despite elevated scrutiny.