OpenAI has launched its OpenAI o1 large language model (LLM), designed to excel in complex reasoning tasks. Accompanying this release is ChatGPT Pro, a premium subscription plan priced at $200 per month. ChatGPT Pro offers users unlimited access to OpenAI o1, o1-mini, GPT-4o, and advanced voice interaction capabilities, making it a comprehensive tool for AI-powered tasks.
The o1-preview version of the model was introduced in ChatGPT on September 12, offering users a glimpse of its capabilities. During this preview, OpenAI o1 demonstrated significant improvements in code generation compared to previous models, highlighting its potential as a robust tool for developers.
The full release of OpenAI o1 was announced on December 5, bringing with it a host of enhancements over its preview version. OpenAI has highlighted that the official release is faster, more accurate, and better equipped for tasks involving coding, math, and reasoning. A key upgrade is the ability to analyze and reason about image uploads, enabling the model to handle visual data more effectively. OpenAI has also optimized o1 for concise and efficient responses, resulting in quicker interactions without compromising on accuracy.
OpenAI’s internal testing underscores o1’s significant performance improvements over its preview version. The model reduces critical errors on challenging real-world questions by 34%, thanks to reinforcement learning techniques that train it to build a detailed internal chain of thought before responding. This advanced reasoning capability allows o1 to excel in both technical and general-use scenarios.
Subscribers to the paid ChatGPT plan can now access o1 through the model selector, replacing o1-preview. Enterprise and education users will gain access to the model next week. Additionally, OpenAI plans to make o1 available in its API, unlocking new capabilities like vision, function calling, developer messages, and structured outputs to facilitate more dynamic interactions with external systems.