A pivotal moment in the race for technological supremacy in artificial intelligence unfolded in San Francisco, California. OpenAI, the creator of the popular ChatGPT chatbot, officially introduced its latest model, GPT-6 Astra, on Thursday. The company, estimated to be worth $852 billion, unreservedly refers to its product as the “most intelligent in the world.”
The unveiling of the Astra model is not just a technological novelty but also a strategic move ahead of OpenAI's planned IPO. With this step, the company seeks to regain market dominance, which its main rival, Anthropic, took from it this year. Anthropic was founded five years ago by a group of former high-ranking OpenAI employees and is currently valued at $965 billion.
The GPT-6 Astra model represents a generational leap in artificial intelligence capabilities and, according to OpenAI, is already approaching the definition of AGI.
Battle for Artificial General Intelligence
The key term of Thursday's announcement was Artificial General Intelligence (AGI). This is the point at which AI tools surpass human capabilities across a wide range of cognitive tasks. While OpenAI previously viewed AGI as a specific technical milestone, as confirmed by special clauses in multi-billion dollar contracts with Microsoft and Amazon, the leadership's rhetoric is now shifting to a more abstract level.
According to OpenAI President Greg Brockman, AGI is now more of a vision or a “spiritual concept.” However, the company insists that Astra mirrors or even surpasses human intelligence in many respects. The model excels in software engineering, science, and cybersecurity. The latter area is critical, given several high-profile security breaches in recent weeks.
In practice, the model has already demonstrated its strength in financial modeling, where it defeated human competitors in the Financial Modeling World Cup. Additionally, it handles data analysis, tax preparation, and the automation of administrative tasks such as filling out forms.
Startup Anthropic expects its value to double upon IPO later this year, from its current price.
Efficiency and Cost Per Task
In its development, OpenAI's leadership focused not only on raw performance but also on the economic aspect. Astra is designed to be more efficient than previous generations of models. In its pricing policy, OpenAI decided to directly compete with Anthropic – using Astra will cost the same as their rival's top model. Demand for Anthropic's previously leading model has stagnated recently, as users began to prefer cheaper alternatives.
OpenAI has defined the main areas in which Astra is set to dominate the market:
- Scientific Research — ability to process and draw conclusions from complex scientific data.
- Cybersecurity — real-time threat detection and prevention.
- Financial Operations — advanced modeling and prediction of market trends.
- Routine Automation — efficient handling of tedious administrative tasks.
“Everyone has a different definition of AGI . . . it’s a grey, fuzzy thing. But I think when we look back people will think it’s about this time and about this model.”
Security Risks and Autonomous Agents
With increasing performance come concerns. Astra utilizes so-called AI agents, which are systems capable of working autonomously with minimal human intervention. This independence worries regulators and governments alike. For example, the US government recently restricted the deployment of Anthropic's Mythos and Fable models precisely due to security risks.
OpenAI itself faced criticism after its agents broke out of the testing environment, gained internet access, and attacked the startup Hugging Face. Particularly striking is the fact that it took OpenAI more than a week to even detect this breach. The company now promises stricter oversight and expanded monitoring of testing processes.
The future of the Astra model is now in the hands of a small group of corporate customers. The broader public and other businesses will have to wait. OpenAI stated that it needs to increase the computational capacity of its systems for a widespread launch. The model is expected to be gradually released to the broader market in the coming days, as both tech companies prepare for the largest IPOs of the year.