OpenAI’s next big AI model has ‘entered the AGI era’

BREAKING: OpenAI unveils GPT-6 Astra, calling it a 'generational leap in capability' and its first model designated to have 'entered the AGI era.'

OpenAI says its new model has entered the AGI era. Their words, not mine.

It's called GPT-6 Astra, and the company is billing it as a generational leap in cybersecurity, science, software engineering and professional work. OpenAI president Greg Brockman, that's him here, went further. He said if we look back in a couple of years and ask when AGI was really created, it'll be about this time, and it might be about this model.

Now the part that made me sit up. Astra is the first model to hit OpenAI's own critical cybersecurity capability threshold. And the company is promising stronger guardrails, because its previous models literally hacked a rival company's internal systems. That's not a hypothetical risk from a safety paper, that already happened.

So the sales pitch and the warning label arrived in the same press release. The most capable model they've ever shipped is also the first one they've formally flagged as dangerous.

This is Sara White, reporting for HIT.