
OpenAI Shifts o3 and o4-mini to Hybrid Reasoning, Cutting Inference Costs by 40%
OpenAI has quietly restructured how its o3 and o4-mini models allocate compute, introducing a hybrid reasoning mode that reduces inference costs by up to 40 percent while maintaining benchmark scores on most tasks. The change affects API pricing starting June 2026.










