Speculation: OpenAI Has a Secret Internal Model Stronger Than GPT-6, But Too Pricey to Ship

haider1 · x · 2026-09-24

An unverified community speculation claims the model Sam Altman expects to call AGI by year-end is a secret internal OpenAI model described as "significantly more capable than GPT-6 astra." The argument: it has already solved 100+ math problems internally, suggesting earlier Navier-Stokes results weren't a fluke—but its lowest reasoning effort reportedly consumes more compute than Astra at maximum, making a public release before year-end very difficult. Unconfirmed rumor.

Original post →

More from Models

Models channel →