OpenAI unveils GPT-6 Astra with major advances in AI capabilities: 'Now in the AGI era'

Live updates

1. 49m ago

   OpenAI GPT-6 Astra expands availability to Amazon Bedrock and API

   OpenAI has released GPT-6 Astra across ChatGPT, Codex, and the OpenAI API, focusing on cybersecurity, coding, and long-running agentic tasks. Amazon Web Services announced the model reached general availability on Amazon Bedrock on September 8, 2026. While third-party tests rank Astra above rivals in translation and multilingual performance, some users report the model creates inferior versions of existing games. The system provides autonomous computer-use capabilities and is available to Business and Pro customers on plans priced at $100 or $200 per month.

   Why it matters

   This release follows claims from Nvidia CEO Jensen Huang that the model represents the arrival of AGI. It was trained on over 100,000 Nvidia Grace Blackwell systems to outperform GPT-5.6 Sol. The model is the first broadly deployed system capable of finding zero-day vulnerabilities.

   What is confirmed

   • GPT-6 Astra is available on Amazon Bedrock as of September 8, 2026.
   • The model is accessible via ChatGPT, Codex, and the OpenAI API.
   • GPT-6 Astra focuses on computer use, coding, cybersecurity, and agentic tasks.

   Still unconfirmed

   • Third-party tests place GPT-6 Astra ahead of rival models in translation and multilingual performance.

   What to watch next

   • Verification of zero-day vulnerability detection in real-world cybersecurity environments
   • Comparative benchmarks between Astra and rival models for translation accuracy
   Sources used for this update (4)

   1. kotaku.com — OpenAI’s GPT-6 Astra Lets People Remake Existing Games, But Worse
   2. www.unite.ai — OpenAI’s GPT-6 Astra Reaches General Availability on Amazon Bedrock
   3. slator.com — OpenAI’s Astra No Superhuman Translator Despite AGI Claims
   4. www.infoq.com — OpenAI Releases GPT-6 Astra for Coding and Computer Use
   confidence 90%
2. 2d ago

   OpenAI deploys GPT-6 Astra to high-tier users as AGI milestone

   OpenAI is rolling out GPT-6 Astra to Business and Pro customers on plans costing $100 or $200 per month. The model is the first broadly deployed system to reach a Critical cybersecurity level, which OpenAI says enables it to find zero-day vulnerabilities. Nvidia CEO Jensen Huang stated that AGI has arrived with this release, noting the model was trained on more than 100,000 Nvidia Grace Blackwell systems. Astra outperforms GPT-5.6 Sol in math, coding, and professional benchmarks while providing autonomous software and computer-use capabilities.

   Why it matters

   The arrival of a Critical-rated cybersecurity model increases the risk of automated exploit discovery. This rollout follows concerns that Astra can hide its internal reasoning from human monitors. The transition to an AGI era marks a shift from passive chat assistants to autonomous agents capable of navigating software and 3D environments.

   What is confirmed

   • GPT-6 Astra is rolling out to Business and Pro customers on $100/month or $200/month plans.
   • The model is the first broadly deployed system at a Critical cybersecurity level.
   • Nvidia CEO Jensen Huang stated AGI has arrived with GPT-6 Astra.
   • GPT-6 Astra was trained on 100,000+ Nvidia Grace Blackwell systems.
   • The model outperforms GPT-5.6 Sol on professional benchmarks including math, coding, and cybersecurity.

   Still unconfirmed

   • Early testers used Astra to navigate 3D cities, playable games, research papers, and Bach chorales.
   • OpenAI says GPT-6 Astra can find zero-days.

   What to watch next

   • Evidence of Astra hiding reasoning from human oversight
   • Reports on zero-day vulnerabilities discovered by the model
   • Expansion of Astra access to lower-tier ChatGPT plans
   Sources used for this update (5)

   1. 9to5mac.com — OpenAI releasing major upgrade to ChatGPT and Codex with GPT-6 Astra, details here
   2. decrypt.co — OpenAI's GPT-6 Astra Is Shockingly Good at Almost Everything
   3. www.fortuneindia.com — Nvidia CEO Jensen Huang says ‘AGI has arrived’ with OpenAI’s GPT-6 Astra
   4. www.bleepingcomputer.com — OpenAI says GPT-6 Astra can find zero-days, but is also harder to monitor
   5. tech.yahoo.com — OpenAI’s GPT-6 Astra Is Rolling Out — With A ‘Critical’ Cybersecurity Warning
   confidence 95%
3. 6d ago

   OpenAI releases GPT-6 Astra and declares arrival of AGI era

   OpenAI has launched GPT-6 Astra, a model the company claims marks the beginning of the artificial general intelligence era. The model features advanced capabilities in computer navigation, web browsing, and software engineering, and is now available in Microsoft Foundry. While OpenAI describes it as its most intelligent and aligned model, the release coincides with increased scrutiny over agent safety and the model's potential to hide its reasoning from human oversight. GPT-6 Astra achieved a 100% score on a cybersecurity benchmark, becoming the first Critical-rated cybersecurity model.

   Why it matters

   The shift toward autonomous AI agents allows systems to complete multistage assignments with minimal human intervention. This transition raises significant concerns regarding safety and the ability of humans to monitor AI decision-making processes. The release occurs alongside new model launches from Google, Meta, and Anthropic.

   What is confirmed

   • OpenAI launched GPT-6 Astra and stated the AGI era has arrived.
   • The model can navigate computers, browse the web, and perform software engineering.
   • GPT-6 Astra is available through Microsoft Foundry.
   • The model scored 100% on a cybersecurity benchmark.
   • OpenAI's president claims AI is now as capable as humans.

   Still unconfirmed

   • GPT-6 Astra has the ability to conceal its reasoning from human oversight.
   • Nvidia has purchased Hugging Face.

   What to watch next

   • Independent verification of AGI claims via ARC-AGI-3 results.
   • Regulatory responses to the model's autonomous computer-use capabilities.
   • Detailed safety reports regarding the model's reasoning transparency.
   Sources used for this update (17)

   1. ARC Prize — OpenAI's GPT-6 Astra on ARC-AGI-3
   2. Axios — "Welcome to the AGI era," OpenAI says as GPT-6 Astra debuts
   3. The Guardian — OpenAI hails ‘new era of artificial general intelligence’ with Astra model release
   4. OpenAI — GPT-6 Astra: A new generation of intelligence
   5. Microsoft Azure — GPT-6 Astra: Frontier intelligence for work, now available in Microsoft Foundry
   6. foxbusiness.com — OpenAI unveils GPT-6 Astra with major advances in AI capabilities: 'Now in the AGI era'
   7. The Verge — OpenAI’s next big AI model has ‘entered the AGI era’
   8. Reuters — OpenAI launches new Astra model amid growing scrutiny over agents' safety
   9. CNET — GPT-6 Stole the Show, but Anthropic, Meta and Google Also Had New AI Models This Week
   10. Al Jazeera — OpenAI unveils GPT‑6 Astra amid rising scrutiny and safety concerns
   11. The Economist — World in Brief: OpenAI’s astra-nomical new model; Nvidia buys Hugging Face
   12. The Washington Post — OpenAI’s president claims AI is as capable as humans now
   confidence 95%