Tech Desk: OpenAI has unveiled its latest frontier model, GPT-6 Astra, boldly declaring it “the most intelligent and aligned model in the world.” This launch comes less than two months after GPT-5.6 Sol, Terra, and Luna, showcasing OpenAI’s rapid development pace despite earlier concerns about slowing model releases.

GPT-6 Astra particularly excels in agentic, computer-use tasks, demonstrating remarkable competence in software engineering, cybersecurity, and general professional work. A compelling demo video highlights its ability to manage multi-step workflows across various domains, from 3D modeling to coding, often simultaneously and without deviation from complex prompts. The model also boasts “strong visual judgment,” critical for complex tasks.

Impressive benchmarks underscore Astra’s capabilities, including a 98.6% score on ARC-AGI-3 for problem-solving and significantly higher scores on Terminal Bench 4.0 for coding and Agent’s Last Exam for agentic tasks compared to its predecessor. Its cybersecurity prowess is particularly noteworthy, achieving a perfect score on ExploitBench, though this raises important questions about responsible deployment.

OpenAI emphasizes Astra’s improved alignment to enhance safety and withstand jailbreak requests, alongside robust monitoring for misuse. Rolling out to limited organizations initially, GPT-6 Astra will soon be available to ChatGPT Plus, Pro, Business, and Enterprise users, as well as via the OpenAI API, priced at $10 per million input tokens and $50 per million output tokens, positioning it as a premium, high-performance tool.