OpenAI o1

OpenAI o1 is a reasoning model from the “o” family, introduced on 12 September 2024 as o1-preview and released as the full o1 model on 5 December 2024. Unlike classic GPT models, o1 is trained with reinforcement learning to generate an extended internal chain of thought before producing an answer.

The model achieves strong results on tasks requiring complex reasoning — mathematics, hard sciences and programming. It scored 83% on AIME 2024 and ranks in the 89th percentile on Codeforces competitive-programming contests. On the GPQA Diamond benchmark it exceeds the accuracy of PhD-level human experts in physics, chemistry and biology.

o1 accepts text and images as input and produces text output. It supports a 200,000-token context window and can generate up to 100,000 tokens in a single response. Its knowledge cutoff is October 2023. It supports function calling, structured outputs and streaming; fine-tuning is not supported.

The model is available commercially through the OpenAI API and Microsoft Azure, as well as in ChatGPT products. API pricing is USD 15 per 1M input tokens, USD 7.50 per 1M cached input tokens and USD 60 per 1M output tokens.