Independent newsroom

Friday, September 4, 2026

The File

Every claim traces to a primary source. Uncertainty is labeled. Corrections are public.

Labs · Friday, September 4, 2026

OpenAI launches GPT-6 Astra and says it is the first of its models to hit a Critical cyber level

The File desk · Sep. 4, 2026, 8:18 a.m. ET

Confirmed6 primary sources

Status

Confirmed as OpenAI’s Sept. 3 launch post (Internet Archive capture; live openai.com returned 403), API docs, and system card. Company benchmark claims are OpenAI’s; not independently verified here.
  • Confirmed
    OpenAI’s Sept. 3, 2026, post “GPT-6 Astra: A new generation of intelligence” introduces GPT-6 Astra. The company says it is rolling out first to a limited set of organizations, then to ChatGPT Plus, Pro, Business, and Enterprise and to the API, Azure, and Bedrock. OpenAI says Astra meets the Critical cybersecurity threshold under its . The API page lists model id gpt-6-astra and Standard pricing of $10 per million input tokens and $50 per million output tokens. The system card, published Sept. 3, says Astra is OpenAI’s first model to reach that Critical cybersecurity level.
  • Unverified
    OpenAI’s benchmark tables (FrontierMath, ARC-AGI-3, ExploitBench, and others) are the company’s. This desk did not independently verify them.
On Thursday, Sept. 3, 2026, OpenAI published “GPT-6 Astra: A new generation of intelligence.” It introduced GPT-6 Astra as its most intelligent and aligned model, with claimed state-of-the-art results on computer use, browsing, software engineering, cybersecurity, science, and professional work. The post says Astra is rolling out first to a limited set of organizations and, over coming days, to ChatGPT Plus, Pro, Business, and Enterprise users, plus the OpenAI API, Microsoft Azure, and AWS Bedrock. The API model id is gpt-6-astra. Standard pricing is $10 per million input tokens and $50 per million output tokens. OpenAI’s system card says Astra is its first model to reach the Critical cybersecurity capability threshold under its . The launch post says the version launching that day will refuse more advanced cybersecurity tasks such as creating proof-of-concept exploits, with less-restrictive defensive access planned through OpenAI Daybreak. Anthropic’s newsroom still led with Sept. 1 Claude Fable 5.1 and Claude Mythos 5.1. Google’s Gemini 3.8 Flash remains Thursday’s File AI item. Live openai.com returned 403 to this desk; the launch-post text is from an Internet Archive capture.
OpenAI published the launch on Thursday. This desk could not read the live page. An Internet Archive capture of openai.com/index/gpt-6-astra/ (snapshot 20260904072121) opens: “We’re introducing GPT‑6 Astra, the world’s most intelligent and aligned model.” “Astra is state-of-the-art on computer use, browsing, software engineering, cybersecurity, science, and professional work.” “GPT‑6 Astra is rolling out today to a limited set of organizations and over the coming days will become available to all ChatGPT Plus, Pro, Business, and Enterprise users, as well as through the OpenAI API, Microsoft Azure, and AWS Bedrock.” The live URL returned HTTP 403 on this desk’s Friday morning fetch.
The same capture: “For developers, GPT‑6 Astra will be available in the OpenAI API as gpt-6-astra and through Microsoft Azure and Amazon Bedrock. OpenAI API Standard pricing is $10 per million input tokens and $50 per million output tokens.” OpenAI’s developers page, fetched live, matches that model id and Standard price ($10 input / $50 output per 1 million tokens) and says Astra “is rolling out today for enterprises in our Trusted Access Program, with access through API and our Plus, Pro, Business and Enterprise plans coming in the coming days.”
On cybersecurity, the launch post says Astra “meets the Critical threshold in cybersecurity under our .” It writes that with the version launching that day, “defenders can use it to complete tasks such as secure code review and patching.” “However, Astra will refuse to comply with more advanced cybersecurity tasks such as creating proof-of-concept exploits for vulnerabilities. Through OpenAI Daybreak, we plan to expand access and roll out less restrictive safeguards in the coming weeks.” This file does not restate OpenAI’s exploit-evaluation tables beyond labeling them company claims.
The GPT-6 Astra System Card on deploymentsafety.openai.com, published September 3, 2026, opens its safety overview: “Today, we are releasing GPT-6 Astra, the most capable model we have ever broadly deployed. Astra is our first model to reach the Critical level of cybersecurity capability under our .” The live safety-overview URL on openai.com returned 403; the same Critical-threshold wording is in this system card. Company eval numbers are OpenAI’s.
Anthropic’s newsroom, fetched Sept. 4, still leads with “Introducing Claude Fable 5.1 and Claude Mythos 5.1,” dated Sept. 1, 2026. Thursday’s File already carried Google’s Gemini 3.8 Flash launch.
How we know

What is still unknown or disputed

  • Company eval numbers are OpenAI’s; not independently verified here.
  • Live openai.com/index/gpt-6-astra/ and openai.com/index/safety-overview-gpt-6-astra/ returned HTTP 403 on this desk pass.
  • Fairwind-style Daybreak acceptance criteria beyond the launch-post and system-card summary were not separately fetched.

Primary sources

Every claim in this story is drawn from the documents below. If a fetch failed, that is recorded on the card.

  1. Source 1 · fetch incomplete

    GPT-6 Astra: A new generation of intelligence

    OpenAI · September 3, 2026

    Live page returned HTTP 403 to this desk. See the Internet Archive capture of the same URL.

    https://openai.com/index/gpt-6-astra/

  2. Source 2

    GPT-6 Astra: A new generation of intelligence (Internet Archive capture)

    Internet Archive / OpenAI · September 3, 2026

    We’re introducing GPT‑6 Astra, the world’s most intelligent and aligned model. Astra is state-of-the-art on computer use, browsing, software engineering, cybersecurity, science, and professional work. GPT‑6 Astra is rolling out today to a limited set of organizations and over the coming days will become available to all ChatGPT Plus, Pro, Business, and Enterprise users, as well as through the OpenAI API, Microsoft Azure, and AWS Bedrock. OpenAI API Standard pricing is $10 per million input tokens and $50 per million output tokens. Astra meets the Critical threshold in cybersecurity under our . However, Astra will refuse to comply with more advanced cybersecurity tasks such as creating proof-of-concept exploits for vulnerabilities. Through OpenAI Daybreak, we plan to expand access and roll out less restrictive safeguards in the coming weeks.

    https://web.archive.org/web/20260904072121/https://openai.com/index/gpt-6-astra/

  3. Source 3 · fetch incomplete

    Safety overview: GPT-6 Astra

    OpenAI · September 3, 2026

    Live page returned HTTP 403. The system card’s opening safety overview states Astra is OpenAI’s first model to reach the Critical cybersecurity capability under its .

    https://openai.com/index/safety-overview-gpt-6-astra/

  4. Source 4

    GPT-6 Astra System Card

    OpenAI Deployment Safety Hub · September 3, 2026

    Published September 3, 2026. Today, we are releasing GPT-6 Astra, the most capable model we have ever broadly deployed. Astra is our first model to reach the Critical level of cybersecurity capability under our .

    https://deploymentsafety.openai.com/gpt-6-astra

  5. Source 5

    GPT-6 Astra Model | OpenAI API

    OpenAI Developers · September 3, 2026

    GPT‑6 Astra is rolling out today for enterprises in our Trusted Access Program, with access through API and our Plus, Pro, Business and Enterprise plans coming in the coming days. Model id gpt-6-astra. Price $10 • $50 Input • Output. Per 1M tokens Input $10.00.

    https://developers.openai.com/api/docs/models/gpt-6-astra

  6. Source 6

    Anthropic Newsroom (checked)

    Anthropic · September 4, 2026

    Top announcement remains Introducing Claude Fable 5.1 and Claude Mythos 5.1 (Sep 1, 2026).

    https://www.anthropic.com/news