DistantNews
Support us
🇪🇸 Spain /Technology

Does OpenAI’s GPT-6 Astra herald a new era for artificial intelligence?

From La Vanguardia · () Spanish

Translated from Spanish and summarized by DistantNews. Read the original for the full story.

At a glance

News Official statement New plan
  • OpenAI announced GPT-6 Astra, which the company says achieved record results in science, mathematics, computer use, programming and cybersecurity tests.
  • Company president Greg Brockman said he personally believes artificial general intelligence may already have been reached.
  • OpenAI delayed the model’s release to adjust its safety measures and says Astra will refuse some cybersecurity tasks.

OpenAI has announced GPT-6 Astra, its most advanced artificial intelligence model, prompting company president Greg Brockman to say that the industry may already have reached artificial general intelligence.

Asked during the presentation whether Astra could qualify as AGI, meaning an AI capable of matching humans across intellectual tasks, Brockman replied: “In my personal opinion, I think we have already reached that point.” He later said it was “not unreasonable to think that we are now in the era of general AI,” describing Astra as only the beginning.

In my personal opinion, I think we have already reached that point.

· Greg BrockmanHe was asked whether GPT-6 Astra should be considered artificial general intelligence.

The model set new records across science, mathematics, computer use, programming, cybersecurity and other areas, according to the article. In ARC-AGI-3, a test designed to assess whether an AI can handle entirely new situations and learn what to do on the fly, Astra scored 99.9%. It also achieved a top score of 98% on FrontierMath Level 4. Brockman said, “I feel that we have experienced a qualitative change.”

It is not unreasonable to think that we are now in the era of general AI.

· Greg BrockmanBrockman described Astra’s performance as evidence that the industry may have entered the AGI era.

The announcement also highlights the limits OpenAI is placing on the system. Astra was classified as being within a critical cybersecurity capability threshold, so OpenAI paused its release while it adjusted safety controls. The company says the model will refuse requests for cybersecurity work, including detecting vulnerabilities in computer systems. OpenAI has described GPT-5 as the model most aligned with its framework for human oversight.

OpenAI says Astra improves speed, accuracy and safety in computer use. It can fill in online forms, update and organize calendars, search the internet, summarize emails, analyze scientific data, create charts, build webpages and perform quality-control checks. Yet the company’s evaluations found that Astra’s written reasoning was harder to supervise than that of its predecessor, GPT-5.6 Sol. Researchers attributed this to greater control over written reasoning in simpler tasks and the ability to solve problems with fewer written steps, while saying the model still struggles to conceal the reasoning needed for complex tasks.

I feel that we have experienced a qualitative change.

· Greg BrockmanHe characterized the shift represented by Astra.
About this summary

Originally published by La Vanguardia in Spanish. Translated, summarized, and contextualized automatically by DistantNews, with a note on how the source frames the story. Not individually reviewed before publishing. How this works.

More Stories

From Our Blog