Is GPT-6 Astra AGI? OpenAI launches ‘world’s most intelligent’ model after 100% score

An interesting turn of events took place Thursday morning, with major artificial intelligence companies reportedly experiencing service disruptions, according to Downdetector. OpenAI, Claude, Grok, Copilot and Cursor were among the services that saw disruption reports. Their respective status pages also showed issues during the day. Gemini users reported similar problems, although Google did not confirm an outage. Microsoft’s Azure cloud platform also recorded a spike in reported problems. The timing created an interesting wrinkle. Several major AI services appeared to stumble on the same day OpenAI began deploying its newest model, GPT-6 Astra. That coincidence quickly fueled speculation online. Some users wondered whether Astra’s deployment had somehow contributed to the disruptions across the broader AI ecosystem. Conspiracy theories remain conspiracies for a reason. There is currently no evidence connecting Astra to the outages. The simultaneous disruptions could stem from entirely unrelated infrastructure problems. The more interesting question sits elsewhere. OpenAI says Astra can handle tasks that increasingly resemble the capabilities associated with artificial general intelligence, or AGI. Perfect cybersecurity score GPT-6 Astra arrives with an unusually strong collection of benchmark results. One figure stands out immediately: a 100% score on ExploitBench, a benchmark focused on cybersecurity capabilities. A perfect score does not automatically prove that a system has achieved AGI. Benchmarks measure defined tasks, not intelligence as a single universal property. The result still matters because cybersecurity demands several forms of technical reasoning. A capable system must understand software, identify weaknesses, and reason about how systems behave under attack. Astra also reportedly scored 98% on FrontierMath Tier 4. OpenAI says the model has already contributed to solving difficult, long-standing mathematical problems. The model reportedly reached 99.9% on ARC-AGI-3 as well. That benchmark targets increasingly difficult forms of abstract reasoning. Taken together, those results create a broader question about what comes next. Is Astra edging toward AGI? OpenAI has not simply built Astra around one narrow capability. The company says the model can perform advanced software engineering, computer use, browsing, and scientific work. That combination makes the AGI debate harder to dismiss as a distant thought experiment. A system that performs strongly across unrelated technical domains looks very different from an AI optimized for one task. Still, benchmark saturation has limits. A perfect score can show that a model has mastered the tested environment. It cannot establish that the system possesses human-like general intelligence. That distinction becomes important as benchmarks approach their ceilings. Once a model reaches near-perfect performance, researchers need harder tests to separate genuine generalization from increasingly sophisticated task optimization. Astra could therefore represent something more interesting than another generational model upgrade. Its benchmark results may offer an early glimpse of where current AI systems begin running into the boundaries of conventional testing. OpenAI’s rollout will put those claims under greater scrutiny. As more users test Astra across real-world tasks, its performance outside controlled benchmarks will matter just as much. If this is not AGI, then what exactly is it? Get the latest in engineering, tech, space & science - delivered daily to your inbox.Aamir is a seasoned tech journalist with experience at Exhibit Magazine, Republic World, and PR Newswire. With a deep love for all things tech and science, he has spent years decoding the latest innovations and exploring how they shape industries, lifestyles, and the future of humanity.