OpenAI has introduced GPT-6 Astra – the first model of the Astra class, which demonstrated an extraordinary performance leap compared to the company's previous generation of models. The new model achieved a score of 100% on the ExploitBench benchmark, indicating a maximum result in testing. The GPT-6 Astra launch is taking place in a limited format: access to the model is being granted only to a select category of clients, implying a gradual rollout and feedback collection prior to a wide release. At a closed press briefing, OpenAI co-founder and President Greg Brockman presented the new model's results to journalists and analysts, emphasizing the significance of the progress achieved in the field of artificial intelligence.
Greg Brockman officially announced the beginning of the AGI (Artificial General Intelligence) era – a general artificial intelligence capable of solving human-level tasks across a wide range of domains. Such a statement from OpenAI's leadership reflects the company's confidence that the technological breakthrough embodied in GPT-6 Astra represents a qualitative leap in the development of AI systems. The perfect score on ExploitBench — a standard benchmark suite for evaluating models' ability to identify vulnerabilities and solve complex tasks – signals that the new model significantly outperforms its predecessors on key metrics. This event has become a landmark moment in the company's history and in the development of generative AI as a whole, drawing the attention of both industry specialists and investors who are closely monitoring innovations in the artificial intelligence sector.
The launch of GPT-6 Astra has had a notable impact on stock markets, particularly on shares of Chinese AI companies, which rallied following the announcement of the new model. This suggests that investors view OpenAI's achievement as a positive signal for the entire artificial intelligence sector, including Chinese AI solution developers. The limited launch format implies that data on GPT-6 Astra's performance in real-world conditions will be accumulated in the near term, enabling the company to optimize the system ahead of a broader deployment. This approach reflects a strategy of responsible rollout for powerful AI systems, given the growing attention from regulators and the public toward issues of safety and ethics in artificial intelligence.