GPT-5.6 SOL is challenged for "mention." OpenAI claims that the reasoning budget adjustment is an experiment

OpenAI responded to GPT-5.6 SOL's question about “mentional” by arguing that the reason for the budget adjustment was simply an experiment. Users found that the internal reasoning budget for the GPT-5.6 SOL tranches fell significantly, from 960 to 128, and suspected that OpenAI had weakened the model in order to save money. Tibo Sottiaux, head of OpenAI Codex, stated that the team had adjusted the reasoning to check the reasons for the fast-consuming subscriptions. The experiment was withdrawn and the official confirmation was not long-term weakening of the model. In addition, OpenAI has completed the optimization of its reasoning, with GPT-5.6 SOL subscriptions receiving approximately 10 per cent more。