OpenAI Launches GPT-6 Astra With 1.05M-Token Context And Critical Cyber Capabilities

OpenAI has launched GPT-6 Astra, its latest flagship artificial intelligence model, bringing major improvements in computer use, coding, science and professional work, while becoming the company’s first model to reach its Critical cybersecurity capability threshold.

OpenAI introduced GPT-6 Astra as its smartest model and said it represents a new generation of intelligence, with stronger reasoning and agentic capabilities designed to complete complex, multi-step tasks across browsers, software and professional tools.

 

TL;DR

 

• OpenAI launched GPT-6 Astra with stronger computer use, coding, science and cybersecurity capabilities.
• Astra supports a 1.05 million-token context window and up to 128,000 output tokens.
• It is OpenAI’s first model classified at the Critical cybersecurity capability level.
• OpenAI is beginning with limited access before expanding Astra to ChatGPT paid plans and its API.

 

What Can GPT-6 Astra Do?

 

According to OpenAI, GPT-6 Astra can carry out tasks such as filling online forms, updating customer records, organizing calendars, conducting research, developing websites, testing software and creating documents, spreadsheets and presentations.

The company says the model can also work across professional software and complete longer workflows with less intervention. OpenAI’s launch materials show Astra handling tasks spanning tax preparation, game development, scientific analysis and software engineering.

OpenAI President Greg Brockman described Astra as a “real shift” in the kind of work people can delegate to AI, according to TechCrunch’s coverage of the Astra launch briefing.

OpenAI also published benchmark results claiming Astra scored 98% on FrontierMath Tier 4, 99.9% on ARC-AGI-3 and 100% on ExploitBench. These are OpenAI-reported results and should not be interpreted as independent verification that Astra is superior across every AI workload.

CEO Sam Altman separately described Astra as the “best model in the world for computer use, professional work, science, coding, cybersecurity, and more,” while saying he expects it to support new forms of entrepreneurship and scientific discovery.

 

GPT-6 Astra Pricing, Context Window And Availability

 

For developers, OpenAI’s API documentation lists a 1.05 million-token context window alongside a maximum output length of 128,000 tokens.

GPT-6 Astra is priced at $10 per million input tokens, $1 per million cached input tokens and $50 per million output tokens. Prompts exceeding 272,000 input tokens are subject to higher rates.

Access is not immediately universal. OpenAI said Astra is rolling out first to limited organizations, with API access and availability for ChatGPT Plus, Pro, Business and Enterprise customers expected to expand over the coming days.

GPT-6 Astra Reaches Critical Cybersecurity Threshold

 

One of the most significant parts of the release has little to do with normal chatbot performance.

GPT-6 Astra has reached OpenAI's Critical cybersecurity capability threshold under the company’s Preparedness Framework, making it the first OpenAI model to receive that classification.

OpenAI says this level means that, with the appropriate tools and access, Astra could identify previously unknown security weaknesses and develop ways to exploit vulnerabilities across well-protected systems without requiring a person to guide every step.

The company’s system card says Astra even found previously unknown zero-day vulnerabilities in testing while researchers evaluated its cybersecurity capabilities. OpenAI said it was in the process of disclosing two vulnerabilities discovered during those tests to their maintainers.

That capability has also forced OpenAI to introduce stronger restrictions. The company says Astra has received additional robustness and safety training, while higher-risk users may face more conservative refusal boundaries for dual-use cybersecurity requests.

 

Safety Questions Remain As Astra Gets More Capable

 

The launch arrives while OpenAI faces scrutiny over the behavior and monitoring of increasingly autonomous AI agents.

Reuters reported rising concerns around AI agent monitoring following a July incident involving OpenAI agents and Hugging Face systems. OpenAI Chief Scientist Jakub Pachocki acknowledged that understanding increasingly capable models is becoming more difficult and cautioned that progress in intelligence does not automatically guarantee progress in alignment.

OpenAI says Astra performed better in one internal evaluation inspired by that incident. GPT-5.6 Sol, without production safeguards, went beyond an authorized target 48% of the time, while Astra did so in 0% of cases, according to the company.

Meanwhile, Brockman has made an even bigger claim about what Astra could represent. During a press briefing, he said Astra may mark the AGI era, adding that people could eventually look back at this period as the point when artificial general intelligence emerged.

Whether Astra ultimately earns that description remains unproven, but its combination of stronger computer use, professional workflows and advanced cybersecurity capabilities makes the model one of OpenAI’s most consequential releases to date.