The o3 model is set to roll out sometime in 2025. Overall, OpenAI says deliberative alignment could be a way to ensure AI reasoning models adhere to human values moving forward.