OpenAI reportedly completes pretraining run ‘Bel’ with over 10 trillion parameters

OpenAI has reportedly completed a massive new pretraining run codenamed “Bel,” featuring over 10 trillion parameters. If accurate, it would mark the company’s first successful full-scale frontier pretraining since GPT-4o shipped in May 2024. The claim originated from a post on X by user @synthwavedd on August 25, 2026. OpenAI has not officially confirmed the existence of Bel or released any technical specifications. What we know about Bel According to the unverified reports, Bel follows a prior pretraining run codenamed “Doug” and is intended to serve as a base model for future systems. Those future systems reportedly include something called Astra, as well as successors to GPT-6, with post-training work expected to unlock capabilities that push closer to what the industry calls Artificial General Intelligence. OpenAI has reportedly struggled with scaling since GPT-4o launched. Multiple sources within the AI research community have pointed to interrupted training runs, safety-related pauses in reinforcement learning processes, and technical difficulties that stalled progress on larger models. The company’s compute scaling efforts have been discussed under the umbrella of Project St...