OpenAI Rolls Out “Budget” Flex Processing Model to Counter Gemini’s Rise

OpenAI has recently launched a new API service called Flex Processing. This “lite” version of the model features slower response times and occasional unavailability, but comes at only half the cost of traditional models. Currently in the testing phase, Flex Processing supports the newly released o3 and o4-mini inference models, and is primarily designed for lower-priority tasks such as model evaluation and data generation. This move is part of OpenAI’s strategy to stay competitive in an increasingly crowded market—particularly in response to Google’s release of the Gemini 2.5 Flash model, which offers strong performance at a more competitive price.