GPT-6 Astra Pro

OpenAI began rolling out GPT-6 Astra on September 3, 2026, framing it in its safety overview as "the most capable model we have ever broadly deployed" and the first OpenAI model to reach the "Critical" tier of cybersecurity capability under its Preparedness Framework. The launch expanded across ChatGPT Plus, Pro, Busin Subsequent to the initial wave, OpenAI confirmed wider rollout to all paid ChatGPT tiers went fully live on September 10, 2026, and separately announced a higher-tier GPT-6 Astra Pro variant reserved for Pro, Business, and Enterprise users, with pricing still to be disclosed. This tier sits within the broader Astra mod

The emergent.sh excerpt confirms GPT-6 Astra launched on September 3, 2026, with phased availability. Enterprise customers in OpenAI's Daybreak cybersecurity program received access first, followed by the OpenAI API, Amazon Web Services, and ChatGPT Plus, Pro, Business, and Enterprise plans "in the coming days." OpenAI For developers, the model is callable as "gpt-6-astra" on the standard OpenAI API at $10 per million input tokens and $50 per million output tokens, with separate cache read/write rates and higher billing multiples for prompts above 272K input tokens. A Fast mode offers twice the speed at twice the standard price, and

According to the Times of India excerpt, OpenAI released GPT-6 Astra with a focus on computer use, software engineering, science, cybersecurity, and professional work, calling it the company's most intelligent and aligned model. It is rolling out first to a limited group of organizations before becoming available to Ch The excerpt reports Astra scored 72.6% on OSWorld 2.0 completing tasks in about 40 minutes, compared with 65.7% and roughly 75 minutes for GPT-5.6 Sol. On Mind2Web, Astra completed tasks 1.9 times faster when combined with an updated Codex system. The article also briefly notes OpenAI introduced GPT-6 Astra Pro for sel

OpenAI rolled out GPT-6 Astra on September 3, 2026, describing it as its most intelligent and aligned model and the first to trigger advanced internal safety protections given its cyber capabilities. Astra reports 98% on FrontierMath Tier 4, 99.9% on ARC-AGI-3 and 100% on ExploitBench, and OpenAI says the model has alr On alignment, OpenAI introduced a new evaluation informed by the Hugging Face incident that measures whether a model will exceed its authorized scope on a hard or impossible task. Without production safeguards, GPT-5.6 Sol went beyond the authorized target in 48% of cases while GPT-6 Astra did so in 0% of cases. Inside

Independent benchmark analysis separates OpenAI's vendor numbers from third-party results. On the Artificial Analysis Intelligence Index, Astra scores 61.2 — effectively tied with predecessor GPT-5.6 Sol and behind Claude Fable 5.1's 65.7 — while OpenAI's own tables show it saturating FrontierMath Tier 4 (97.6%), ARC-A Astra ties Fable 5.1 on the neutral coding index and trails only on Artificial Analysis's Claude Code setup, but loses Humanity's Last Exam to Fable 5.1, so the launch is not a clean sweep. The real practical gain is coding efficiency: Astra matches Fable 5's coding-agent score at less than half the cost per task and c

OpenAI began rolling out GPT-6 Astra on September 3, 2026 to a limited set of customers, framing it as a leap in autonomous computer use rather than chat. Astra is built to operate software the way a person does — navigating browsers, spreadsheets, websites and desktop apps to produce finished documents and carry out m At the launch briefing, president Greg Brockman closed with "Welcome to the AGI era," but hedged the definition: he said he personally believes OpenAI has reached AGI while leaving users to decide whether Astra meets the bar, and noted the term is no longer tied to its earlier contractual trigger with Microsoft — descr

OpenAI launched GPT-6 Astra on September 3, 2026 as its new frontier flagship, succeeding GPT-5.6 Sol and priced at $10 per million input tokens and $50 per million output tokens. The model is positioned around state-of-the-art computer use, professional-work acceleration, and a jump in cybersecurity capability that cr On OpenAI's own evaluations, Astra hits 97.6% on FrontierMath Tier 4, 100% on ExploitBench, 72.6% on OSWorld 2.0 at roughly 47% less time per task than Sol, and 99.9% on ARC-AGI-3 — though DataCamp flags that the ARC-AGI-3 saturation depends on a stateful, expensive provider-adapter harness and drops sharply under stat