OpenAI: o1-pro

OpenAI o1-pro is a text model from OpenAI, part of the o1 series. OpenAI describes the o1 family as models trained with reinforcement learning to think before they answer and to perform complex reasoning, and o1-pro is the variant that uses more compute to think harder and provide its answers. It sits in the text category and was released by the provider on 2025-03-19.

The model accepts text, images and files as input, and it returns text. Its context window is 200,000 tokens, and the maximum output is 100,000 tokens. The knowledge cutoff is 2023-10-31. Because it reasons before responding, responses may take longer and consume more tokens than a standard chat model, which is worth keeping in mind when you design around latency and output budgets.

The combination of a large context window, file and image input, and deliberate reasoning points to work where the model needs to digest a lot of material before answering. Examples include analysing long documents or codebases supplied as files, working through multi-step problems where intermediate reasoning matters, extracting structured results from mixed text and image inputs, and tasks that benefit from the model being able to examine a problem from several angles before committing to an answer. The 200,000 token context makes it practical to pass substantial source material in a single request rather than chunking it.

You call o1-pro through api14 using the OpenAI-compatible API with the model id "openai/o1-pro". The supported request parameters are include_reasoning, max_tokens, reasoning, response_format, seed and structured_outputs. include_reasoning and reasoning relate to the model's thinking behaviour, response_format and structured_outputs let you constrain the shape of the returned text, max_tokens caps the output, and seed supports reproducible sampling where the backend honours it.

Because reasoning happens before the answer, o1-pro is not the obvious choice for simple, latency-sensitive chat where a lighter text model would do. The knowledge cutoff of 2023-10-31 also means the model has no awareness of events after that date, so anything time-sensitive needs to be supplied in the prompt or retrieved separately. Output is text only, so if you need generated images or audio, you will need a different model.