OpenAI Declares, "The AGI Era Has Begun"—Counterattacks Anthropic with GPT-6 Astra

【Financial News New York = Reporter Lee Byung-chul】  OpenAI has declared the beginning of the artificial general intelligence (AGI) era by unveiling a new artificial intelligence (AI) model, GPT-6 Astra. OpenAI, which has been seen as losing its technological lead to its biggest rival, Anthropic, this year, has launched a counterattack with its next-generation model.
OpenAI unveiled GPT-6 Astra on the 3rd (local time) and introduced it as "the most intelligent model in the world." OpenAI, valued at $852 billion, said Astra delivers market-leading performance in software engineering, science, cybersecurity, and other fields.
OpenAI placed particular emphasis on AGI. AGI refers to AI capable of demonstrating abilities comparable to or greater than those of humans across various cognitive domains.
OpenAI President Greg Brockman described Astra as having achieved "a generational leap in performance." He said, "Everyone has a different definition of AGI, and it is an ambiguous concept," adding, "I think that when people look back in the future, they will think of this period and this model."
The remarks effectively expressed confidence that Astra could be regarded as the model that ushered in the AGI era.
OpenAI’s public reference to the arrival of AGI is significant. The company has long defined AGI as a specific milestone to be reached in the AI development process and has included so-called "AGI clauses" in major investment agreements. Brockman explained, however, that AGI is now "closer to a mission or a spiritual concept" than a specific technical standard.
The launch of Astra also represents a counterattack against Anthropic.
OpenAI effectively pioneered the generative AI market by launching ChatGPT in late 2022. This year, however, it has been widely viewed as having lost its technological edge to Anthropic. Founded by former OpenAI employees, Anthropic has seen its valuation soar to as much as $965 billion, surpassing OpenAI’s $852 billion.
As both companies pursue initial public offerings, the technology race is also expanding into a battle over corporate valuations. Anthropic could reportedly receive a valuation of up to twice its current level in an IPO process later this year. OpenAI is also pursuing a listing, making Astra’s performance and its ability to secure corporate customers key factors that are expected to determine its future valuation.
The price competition is also intensifying. Astra is expected to be priced at a level similar to Anthropic’s highest-performing model. OpenAI emphasized instead that Astra is more efficient than its previous models.
Brockman said, "What matters is the cost per task," adding, "The key is whether you can complete the task you want at an appropriate price and speed."
OpenAI explained that Astra also demonstrated strong performance in corporate tasks such as financial modeling, tax-filing preparation, and data analysis. The company said Astra outperformed human participants in a financial modeling competition and could handle repetitive tasks such as preparing various documents. The center of gravity in the AI race is rapidly shifting from simple question-and-answer interactions to "AI agents" that independently perform real-world tasks.
However, as AI capabilities approach the realm of human abilities, concerns about safety are also growing.
OpenAI assessed that Astra had reached the "Critical" level of cybersecurity capability for the first time under its safety standards. Given the appropriate tools and access permissions, Astra could identify previously unknown security vulnerabilities and even develop methods to exploit them without step-by-step instructions from a person. Accordingly, OpenAI plans to initially provide Astra to a small number of companies, review cybersecurity issues, and expand access over the coming days.



[email protected] Reporter Lee Byung-chul Reporter