OpenAI o1 is a new large language model (LLM) trained with reinforcement learning to perform complex reasoning.
AI research and deployment company, OpenAI has introduced a new series of AI models, OpenAI o1 and OpenAI o1-mini. These AI models are designed to spend more time thinking before they respond.
The o1-preview possesses strong reasoning capabilities and broad world knowledge. The o1-mini, on the other hand, provides a faster, 80% cheaper, and more competitive solution for coding tasks.
According to OpenAI, “We’ve developed a new series of AI models designed to spend more time thinking before they respond. They can reason through complex tasks and solve harder problems than previous models in science, coding, and math.”
OpenAI o1: A New Series of Reasoning Models
The o1-preview is the first of this series in ChatGPT and our API. This is a preview and the company expects regular updates and improvements.
ChatGPT Plus and Team users will have access to o1 models as of right now. The model selector allows you to manually choose between o1-preview and o1-mini. The weekly rate limits at launch will be 30 messages for o1-preview and 50 for o1-mini. OpenAI aims to raise these rates and make ChatGPT capable of selecting the appropriate model on its own for each request.
Alongside this release, the company also includes evaluations for the next update, currently in development. The new LLMs currently lack several features offered by that model, notably the ability to assess files uploaded by the user. There are also no integrations that would allow o1 to interact with external applications.
However, the new LLM series is significantly better at tasks that require reasoning skills.
OpenAI developers had o1-preview finish a qualifying exam for the U.S. Math Olympiad in one internal test. The model performed significantly better than GPT-4o, with average scores ranging from 74% to 93%. According to OpenAI, o1-preview’s highest average score placed it among the top 500 test-takers in the United States.
The GPQA Diamond benchmark is a set of challenging science questions that the ChatGPT developer had o1-preview attempt to answer in a different assessment. In a series of questions covering physics, biology, and chemistry, the model outperformed a group of doctorate-holding specialists.
The company claims that o1’s usage of a machine learning technique called CoT, or chain of thought, is one factor in its superior thinking abilities. Using this strategy, LLMs can take a big task and divide it into smaller pieces, which they can then complete one at a time. This approach can often assist an LLM increase the accuracy of its responses when dealing with difficult requests.
The OpenAI o1 series is accessible through ChatGPT as well as through its API, which enables programmers to incorporate its LLMs into their applications.
Stay Tuned to The Future Talk for more AI news and insights!