OpenAI released o1, a new model with improved performance on long-horizon tasks. The model is available in a limited preview for select developers. You can apply for access to test o1. The model targets applications requiring multi-step reasoning and planning.