OpenAI has taken back the lead from Anthropic with its latest AI model, Astra, after losing ground to its rival in the race to build the most capable artificial intelligence.
The company launched GPT-6 Astra on Thursday, describing it as its most intelligent model so far and a major jump over its previous systems.
OpenAI President Greg Brockman went even further, saying Astra could mark the beginning of the artificial general intelligence era: the long-discussed point where AI becomes capable of doing most economically valuable work at a human level.
The claim comes at an important time for OpenAI.
Anthropic has been gaining ground with its Claude models, particularly among businesses and professional users. Its latest Fable 5.1 and Mythos 5.1 models were released just days before Astra.
OpenAI says Astra is better at a wide range of jobs, including software engineering, scientific work, financial modelling, tax preparation and data analysis.
But one of its biggest jumps is in cybersecurity.
The model can find previously unknown flaws in computer systems and, in testing, create exploits without a person guiding every step. OpenAI has therefore placed its most powerful cyber capabilities behind additional restrictions and is initially giving them to a limited group of trusted testers.
That ability is also one of the reasons Astra has attracted so much attention.
OpenAI is not treating the model like an ordinary software update. The company says its stronger capabilities make it harder to monitor and require new safety measures.
OpenAI and Anthropic are both preparing for potential stock-market listings, making the race over AI customers, revenue and technical leadership even more important.
2026-09-03 23:57:00
