OpenAI Launches GPT-6 Astra as the World's Most Intelligent Model
Astra is designed to operate computers, browse the web, write and test software, conduct research and produce documents, spreadsheets and presentations.
OpenAI has launched GPT-6 Astra, its latest flagship AI model, positioning it as a major step toward AI systems capable of carrying out complex, end-to-end tasks rather than simply responding to prompts.
Astra is designed to operate computers, browse the web, write and test software, conduct research and produce documents, spreadsheets and presentations. OpenAI said the model can perform tasks such as filling online forms, updating CRM records, organising calendars, building websites and troubleshooting software directly on a user's screen.
The Left Shift Bureau
The model marks a broader shift in how OpenAI is positioning its AI systems. Rather than requiring users to provide detailed instructions for individual steps, Astra can execute multi-stage workflows across applications, effectively functioning as an AI operator. OpenAI said it delivers state-of-the-art performance in computer use, browsing, software engineering, cybersecurity, science and professional work.
Astra also posted strong results on several demanding benchmarks. OpenAI said it achieved a 98% score on FrontierMath Tier 4, 99.9% on ARC-AGI-3 and 100% on ExploitBench, while also delivering state-of-the-art results on benchmarks covering computer workflows and scientific tasks.
The company has also focused heavily on speed. OpenAI said an updated Codex harness combined with Astra's capabilities allows tasks on the Mind2Web benchmark to be completed 1.9 times faster than with GPT-5.6 Sol.
The model comes with a 1.05-million-token context window and can generate up to 128,000 output tokens. API pricing starts at $10 per million input tokens and $50 per million output tokens, according to OpenAI's developer documentation.
However, Astra's launch also highlights the growing safety challenges surrounding increasingly capable AI agents. OpenAI said earlier this week that Astra crossed its “Critical” threshold for cybersecurity capability, meaning it can identify previously unknown vulnerabilities and develop exploits across well-protected systems with limited human guidance.
The Left Shift Bureau
As a result, OpenAI has introduced stronger safeguards and is initially restricting its most advanced cybersecurity capabilities to selected testers.
Astra is initially rolling out to organisations in OpenAI's Trusted Access Program, with availability planned for ChatGPT Plus, Pro, Business and Enterprise users and through the API.
The Left Shift Bureau
Pritam Bordoloi
