OpenAI has officially released its OpenAI o1 large language model (LLM), built for complex reasoning. The company also introduced ChatGPT Pro, a $200 monthly plan that includes unlimited access to OpenAI o1, o1-mini, GPT-4o, and advanced voice conversations.
Available in a preview version, called o1-preview, in ChatGPT beginning September 12, OpenAI o1 proved to be better at code generation than previous OpenAI models.
The general release of OpenAI o1 was announced December 5. Compared to o1-preview, users can expect a faster, more powerful, and more accurate reasoning model that is better at coding and math, OpenAI said. A key improvement is that o1 now can be used to reason about image uploads, opening up new applications by explaining and analyzing visual representations with more accuracy and detail. The official o1 release is also more concise in its thinking, OpenAI said, resulting in faster response times than the o1 preview.
OpenAI testing has shown that o1 significantly outperforms o1-preview, reducing major errors on difficult real-world questions by 34%, according to the company. OpenAI o1 was trained with reinforcement learning to perform complex reasoning, and can produce a long internal chain of thought before responding to the user, OpenAI said.
Paid ChatGPT users can access o1 through the model selector, replacing o1-preview. Enterprise and Edu users gain access to o1 next week. Also, o1 will be made available in the API with new capabilities including vision, function calling, developer messages, and structured outputs, to enable richer interactions with external systems, OpenAI said.