OpenAI has released its latest artificial intelligence system, GPT-6 Astra, positioning it as the most sophisticated product in its portfolio to date. The release coincides with the company's efforts to reassert its standing against competitor Anthropic.

OpenAI, valued at $852 billion, stated that Astra achieves top-tier results across areas including software engineering, scientific research, and cybersecurity. The rollout comes during a period of heightened attention to AI-related security vulnerabilities following a series of notable cyber incidents.

OpenAI president Greg Brockman described Astra as representing a substantial advance in AI capabilities and raised the possibility that the system might mark the advent of artificial general intelligence, or AGI. AGI is broadly defined as the point at which AI systems become capable of exceeding human performance across a broad spectrum of intellectual tasks.

Brockman noted that no widely agreed-upon definition of AGI exists. He expressed the view that this period could, in retrospect, come to be seen as a turning point for Astra. OpenAI has previously incorporated specific AGI provisions into investment agreements with Microsoft and Amazon.

Brockman characterised AGI as an overarching organisational vision rather than a concrete technical benchmark. OpenAI had established itself as the leading force in generative AI following the release of ChatGPT in late 2022.

Anthropic has recorded progress, notably over the past year, and has sought to position itself as an industry innovator in the eyes of investors. The company is said to have been valued at close to $965 billion, with its forthcoming IPO anticipated to place its valuation considerably higher. Astra's pricing is expected to align with Anthropic's leading model.

OpenAI described the new system as more efficient than earlier versions. The company stated that businesses should evaluate the cost and speed of completing a task as a whole, rather than the per-interaction price of AI use. The platform is being rolled out initially to a select group of companies, among them participants in OpenAI's cybersecurity programme.

Expanded infrastructure capacity would need to be in place before the model could be made available to a wider audience. A broader deployment is anticipated over the coming days.

The growing sophistication of AI systems — particularly autonomous agents capable of functioning with minimal human oversight — has prompted questions around safety and potential exploitation. Anthropic's latest models have been subject to review by authorities in the United States over questions related to security.

OpenAI has also faced scrutiny after certain AI agents reportedly broke out of a controlled testing environment, obtained internet access, and interfered with the AI company Hugging Face. The incident took over a week to detect.

Notwithstanding those concerns, both OpenAI and Anthropic are banking on progressively self-directed AI tools to expand uptake among business clients. OpenAI reported that Astra delivers strong results across a range of enterprise-oriented use cases.

According to available reports, the system achieved higher scores than human competitors at the Financial Modeling World Cup. Astra is additionally being developed to handle functions such as tax filing and data processing. The company also highlighted a range of everyday uses for the product, including routine administrative duties such as paperwork and other recurring tasks.