OpenAI Announces AGI Era with GPT-6 Astra Launch: Microsoft and Amazon as Major Beneficiaries

OpenAI launched its new AI model GPT-6 Astra on September 3-4, 2026, positioning it as the beginning of the artificial general intelligence (AGI) era. OpenAI co-founder and President Greg Brockman stated at a closed press briefing: "Welcome to the AGI era." The company emphasized, however, that this does not constitute a formal declaration of achieving AGI. Instead, OpenAI described AGI as a "mission concept or spiritual concept" and leaves it to observers to decide for themselves whether Astra has reached this threshold.

Microsoft and Amazon as exclusive cloud distribution partners

GPT-6 Astra will be made available through multiple distribution channels, with Microsoft and Amazon serving as central cloud partners. The model is accessible via Microsoft Azure, Amazon Web Services (AWS) Bedrock, and the OpenAI API. Additionally, users of ChatGPT subscriptions Plus, Pro, Business, and Enterprise will have access to the new model.

A joint statement from OpenAI and Microsoft on September 4, 2026 confirmed: "The AGI definition and processes remain unchanged. The contractual definition of AGI and the process to determine whether it has been achieved remain the same. The partnership supports OpenAI's growth." The reference pertains to a clause in the OpenAI-Microsoft partnership agreement that previously tied certain conditions to an AGI determination. OpenAI stated that "there is no contractual AGI trigger anymore, so this is actually no longer a relevant concept."

Technical capabilities and benchmark results

OpenAI describes Astra as "the world's best computer-use model," developed to operate software like a human—across browsers, spreadsheets, websites, desktop applications, and engineering tools, without requiring dedicated API integrations for each application.

In internal cybersecurity benchmarks conducted between June and August 2026, Astra achieved significantly higher success rates than its predecessor GPT-5.6 Sol. In the ExploitBench test, which evaluates exploit development based on vulnerabilities from June to August 2026, Astra achieved substantially higher rates of arbitrary code execution while using significantly fewer output tokens. During evaluation, Astra discovered and exploited two previously unknown zero-day vulnerabilities, which OpenAI reported to the respective developers.

In the SRE-Bench test, which examines binary reverse-engineering capabilities and software logic comprehension, Astra solved 88.0 percent of tasks on the first attempt and 99.2 percent within four attempts. By comparison, GPT-5.6 Sol achieved 55.9 percent and 68.7 percent, respectively.

Enterprise use cases and practical applications

Astra was developed to handle multi-step workflows and concrete business tasks. Use cases include:

• Filling out online forms and updating CRM records
• Organizing calendars and conducting web research
• Manipulating spreadsheets and analyzing scientific data in Python notebooks
• Working in Power BI
• Creating and testing websites
• Operating engineering applications such as KiCad and FreeCAD
• Installing and troubleshooting software
• Creating finished documents and presentations via voice input

Rollout schedule and access

Enterprise customers gained first access to GPT-6 Astra through OpenAI's staged access program Daybreak, which launched on September 4, 2026. Users of ChatGPT Plus, Pro, Business, and Enterprise subscriptions were to receive access in the following days. Availability via the cloud platforms Azure and AWS Bedrock as well as through the OpenAI API was announced for "the coming days."

OpenAI plans to expand access through the Daybreak program in the coming weeks with less restrictive security measures to enable defensive workflows—including vulnerability and proof-of-concept validation, malware analysis, and detection engineering.

Security measures and pricing

The version of Astra introduced on September 4, 2026 includes multiple security measures: the model refuses to perform advanced cybersecurity tasks such as creating proof-of-concept exploits for vulnerabilities. Additional measures include enhanced robustness against potential jailbreaks, extended context for monitoring systems, and rigorous internal and external testing including automated evaluations with internal red-teaming adversaries.

OpenAI's standard API pricing is $10 per million input tokens and $50 per million output tokens.

Market environment and competition

Concurrent with the Astra launch, there were indications of a slight advantage for one of the two cloud partners, Microsoft or Amazon, without details on the relative benefits being disclosed. Competitor Anthropic released its competing model Fable 5.1, which according to the company achieved more than double the science score of its predecessor. Comparative performance data between Fable 5.1 and Astra was not available at launch time.