o1-preview is our new reasoning model for complex tasks. The model has 128K context and an October 2023 knowledge cutoff.