Microsoft Leak Reveals OpenAI's GPT-6 Uses Secret 'Looped' Trick

Microsoft has confirmed—perhaps by accident—that OpenAI's upcoming GPT-6 series uses a technique called 'looped transformers.' This was spotted on a public Microsoft webpage, which has since been taken down. Looped transformers let an AI model run through its reasoning multiple times, like double-checking its work. The leak also mentioned GPT-6.1 Sol, a version that uses two passes instead of three, making it more efficient.

So what does this mean for you? AI models like GPT-6 power chatbots, writing assistants, and coding helpers. If they can work faster and use less computing power, the cost of running them drops. Those savings could be passed on to you through cheaper subscriptions or free tiers. Also, a more efficient AI means quicker responses—no more waiting for a chatbot to think.

The catch? This is still insider info from a leak, not an official announcement. Microsoft removed the page, so details are unconfirmed. Also, 'looped transformers' is a technical term; in plain English, it's like the AI taking a few extra seconds to think before answering. That extra thinking could make answers more accurate, but it also might make the AI slower if not optimized well.

Overall, this leak gives us a peek into how OpenAI is trying to make GPT-6 smarter without making it too expensive. For everyday users, it hints at a future where AI is both more powerful and more affordable. Keep an eye out for official news from OpenAI, likely later this year.