OpenAI will reportedly ship at least one model this year but its flagship still needs alignment work
OpenAI will reportedly ship one or two models by year-end, but its looped-transformer flagship still needs alignment work and safety test infrastructure.
The Information reports OpenAI employees expect one or two new models before year-end, while the major flagship upgrade still requires alignment work and testing infrastructure. The flagship uses looped transformers ('Recurrent Depth'), the technique behind Astra, which boosts math/coding performance but hides reasoning in internal representations, weakening chain-of-thought monitoring. After the Hugging Face incident, OpenAI moved roughly 10% of compute (hundreds of megawatts) to alignment and safety. A deleted Microsoft write-up accidentally revealed GPT-6.1 Sol uses looping inference with two passes instead of three.