OpenAI Launches GPT-6 Astra: First Model Crossing 'Critical' Cybersecurity Threshold
On Thursday, September 3, 2026, OpenAI officially began rolling out its next-generation frontier flagship model, GPT-6 Astra (gpt-6-astra). Available across ChatGPT Plus, Pro, Business, and Enterprise tiers, as well as via the OpenAI API, Microsoft Azure, and Amazon Bedrock, the model succeeds GPT-5.6 Sol as the industry standard for advanced reasoning, agentic execution, and computer use.
The release marks a historic milestone in AI safety policy: GPT-6 Astra is the first model in OpenAI's history to cross the "Critical" cybersecurity capability threshold under its Preparedness Framework. Consequently, OpenAI has implemented strict deployment safeguards, including restricting access to its most advanced cyber capabilities, deploying active chain-of-thought monitoring, and limiting higher-risk accounts.
Breakthrough Agentic Benchmarks and Computer Use
GPT-6 Astra shifts the frontier focus from raw text generation to autonomous execution and computer control. On the OSWorld 2.0 OS-use benchmark, GPT-6 Astra achieved a score of 72.6%, significantly outperforming GPT-5.6 Sol (65.7%) and Anthropic's Claude Opus 5 (70.2%).
Furthermore, the model features an updated Codex harness, which OpenAI reports delivers 1.9x faster task completion compared to the GPT-5.6 Sol experience on the Mind2Web agent benchmark. In coding, on the FrontierCode 1.1 Extended benchmark, GPT-6 Astra recorded a score of 64.5%, surpassing Claude Fable 5.1's 63.6% and GPT-5.6 Sol's 60.6%.
The release occurs amidst heightened industry scrutiny following the July 2026 Hugging Face breakout incident OpenAI Hugging Face Breakout Incident and the Call for Collective Cyber Defense, pushing OpenAI to implement rigorous pre-release vetting in alignment with the White House's Executive Order 14409 White House Finalizes EO 14409 AI Vetting Framework, Exempting Open-Weight Models in Major Policy Shift.
"GPT-6 Astra brings together years of research and big bets across pre-training, reinforcement learning, and alignment." — OpenAI, "GPT-6 Astra: A new generation of intelligence," September 3, 2026
"OpenAI on Thursday announced it will begin rolling out its latest artificial intelligence model, GPT-6 Astra, which the company said is the product of “years of research and big bets.” OpenAI disclosed earlier this week that Astra is its first model to reach its “Critical” internal cybersecurity threshold, and said it planned to limit access to those advanced capabilities." — CNBC, "OpenAI announces rollout of GPT-6 Astra model," September 3, 2026