{"id":2347,"date":"2026-09-04T08:30:33","date_gmt":"2026-09-04T08:30:33","guid":{"rendered":"https:\/\/techvisor.pro\/?p=2347"},"modified":"2026-09-04T08:30:33","modified_gmt":"2026-09-04T08:30:33","slug":"gpt-6-astra-openais-new-model-hailed-as-a-step-toward-agi","status":"publish","type":"post","link":"https:\/\/techvisor.pro\/en\/gpt-6-astra-openais-new-model-hailed-as-a-step-toward-agi\/","title":{"rendered":"GPT-6 Astra: OpenAI’s new model, hailed as a step toward AGI"},"content":{"rendered":"

On September 3, 2026, OpenAI released GPT-6 Astra \u2014 a model the company calls a \u201cgenerational leap\u201d and the most powerful development in its entire history. We break down what the new model can do and why its release was accompanied by unusual caution from the company itself.<\/p>\n

What Is GPT-6 Astra<\/h2>\n

Astra is OpenAI\u2019s new flagship, which the company describes as \u201cthe smartest and best-aligned model in the world.\u201d According to the company, Astra demonstrates the highest level among all models in working with computers, browsers, software development, cybersecurity, science and professional tasks.<\/p>\n

OpenAI president Greg Brockman said that the model may eventually be seen as the moment when AGI (artificial general intelligence) emerged. According to CEO Sam Altman, Astra has taken him to a \u201cnew level of capabilities\u201d and has already changed his own workflows.<\/p>\n

A Record in Mathematics and New Capabilities<\/h2>\n

\"A<\/p>\n

Astra achieved 98% on FrontierMath Tier 4 \u2014 one of the most difficult mathematics benchmarks, developed specifically to test problems that cannot be solved through memorized solutions. According to OpenAI, the model has already helped solve long-standing open problems in mathematics.<\/p>\n

A separate new feature concerns Codex: instead of the standard \u201ccompression\u201d of conversation history when the context window fills up (when the model simply summarizes previous work), Astra received a new way to preserve and restore context during long sessions \u2014 for example, during complex debugging or large-scale code refactoring.<\/p>\n

Why the Release Was Delayed and What \u201cCritical\u201d Cyber Risk Means<\/h2>\n

\"Why<\/p>\n

The most important nuance of this release is not the capabilities, but the caution with which OpenAI approached it. The company initially planned a public release for September 5, but delayed it after an incident involving a leak on Hugging Face in July 2026 in order to add additional safeguards.<\/p>\n

On September 1, OpenAI officially reported that, based on additional testing, Astra meets the threshold for a \u201cCritical\u201d level of cyber capability under the company\u2019s internal Preparedness Framework \u2014 the first OpenAI model to receive this classification. In simple terms, with the necessary tools and access, the model could theoretically independently, without step-by-step human guidance, find previously unknown vulnerabilities and develop ways to exploit them in well-protected systems.<\/p>\n

Because of this, access to the model\u2019s most advanced cybersecurity capabilities will be significantly restricted. For now, Astra is available only to customers of the Daybreak cybersecurity program \u2014 companies that undergo a separate verification process.<\/p>\n

When Astra Will Become Available to Regular Users<\/h2>\n

\"When<\/p>\n

In the coming days after the initial release, Astra will become available to users of ChatGPT Plus, Pro, Business and Enterprise plans, as well as through the OpenAI API and Amazon Web Services.<\/p>\n

In Brief: The Key Points<\/h2>\n