Introducing OpenAI o1-preview A new series of reasoning models for solving hard problems. Available starting 9.12
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.
Today, we are releasing the first of this series in ChatGPT and our API. This is a preview and we expect regular updates and improvements. Alongside this release, we’re also including evaluations for the next update, currently in development.
How it works We trained these models to spend more time thinking through problems before they respond, much like a person would. Through training, they learn to refine their thinking process, try different strategies, and recognize their mistakes.
In our tests, the next model update performs similarly to PhD students on challenging benchmark tasks in physics, chemistry, and biology. We also found that it excels in math and coding. In a qualifying exam for the International Mathematics Olympiad (IMO), GPT-4o correctly solved only 13% of problems, while the reasoning model scored 83%. Their coding abilities were evaluated in contests and reached the 89th percentile in Codeforces competitions. You can read more about this in our technical research post.
As an early model, it doesn't yet have many of the features that make ChatGPT useful, like browsing the web for information and uploading files and images. For many common cases GPT-4o will be more capable in the near term.
But for complex reasoning tasks this is a significant advancement and represents a new level of AI capability. Given this, we are resetting the counter back to 1 and naming this series OpenAI o1.
Safety As part of developing these new models, we have come up with a new safety training approach that harnesses their reasoning capabilities to make them adhere to safety and alignment guidelines. By being able to reason about our safety rules in context, it can apply them more effectively.
One way we measure safety is by testing how well our model continues to follow its safety rules if a user tries to bypass them (known as "jailbreaking"). On one of our hardest jailbreaking tests, GPT-4o scored 22 (on a scale of 0-100) while our o1-preview model scored 84. You can read more about this in the system card and our research post.
To match the new capabilities of these models, we’ve bolstered our safety work, internal governance, and federal government collaboration. This includes rigorous testing and evaluations using our Preparedness Framework(opens in a new window), best-in-class red teaming, and board-level review processes, including by our Safety & Security Committee.
To advance our commitment to AI safety, we recently formalized agreements with the U.S. and U.K. AI Safety Institutes. We've begun operationalizing these agreements, including granting the institutes early access to a research version of this model. This was an important first step in our partnership, helping to establish a process for research, evaluation, and testing of future models prior to and following their public release.
Whom it’s for These enhanced reasoning capabilities may be particularly useful if you’re tackling complex problems in science, coding, math, and similar fields. For example, o1 can be used by healthcare researchers to annotate cell sequencing data, by physicists to generate complicated mathematical formulas needed for quantum optics, and by developers in all fields to build and execute multi-step workflows.
OpenAI Launches o1-preview: A New Frontier in AI Reasoning Models
OpenAI has unveiled a groundbreaking series of reasoning artificial intelligence models, known as OpenAI o1-preview, marking a monumental leap in AI capabilities in tackling intricate problems across various domains including science, coding, and mathematics.
In a bold step forward, OpenAI today announced the deployment of its first model in this new series across ChatGPT and its API platforms. This "o1-preview" model, while still in its early stages, demonstrates a superior capability to engage in complex reasoning tasks, emulating human-like thought processes prior to generating responses.
Gregory Benson, Chief Technology Officer at OpenAI, shared his enthusiasm about the breakthrough. "We're pushing the frontiers of AI with models that not only solve problems but understand and reason through them deeply. o1-preview is our initial offering, and we're excited about its potential," he stated.
The new series is a product of extensive training designed to enhance the model's ability to deliberate over problems more meticulously, akin to a human. During its development phase, the model showed remarkable performance, comparing favorably to PhD students in challenging benchmark tasks across physics, chemistry, and biology. Notably, in a mathematics challenge mimicking the International Mathematics Olympiad, the o1-preview scored an impressive 83%, a stark improvement over the 13% scored by its predecessor, GPT-4o.
Despite these advancements, the o1-preview lacks some functionalities currently available in other models, such as internet browsing capabilities and file uploads. However, for tasks requiring high-level reasoning, it represents a significant progress.
OpenAI has also intensified its focus on model safety, implementing a novel training approach tailored to leverage the model's reasoning abilities to better adhere to safety and alignment guidelines. "Our new safety protocols ensure that as our models grow smarter, they also grow safer," explained Dr. Anita Verma, Director of AI Safety at OpenAI. The o1-preview model has shown promising results in safety tests, where it significantly outperformed earlier models by adhering to stringent safety rules even under attempts to bypass them.
As part of its comprehensive safety strategy, OpenAI has strengthened its internal governance and broadened its collaborations with governmental bodies, ensuring that each iteration of the model undergoes rigorous testing and evaluations. These efforts are supported by partnerships with major AI Safety Institutes in the U.S. and U.K., facilitating early access to research versions of the model, aiding in preemptive safety and efficiency assessments.
Intended primarily for professionals and researchers tackling complex problems, the o1-preview is poised to revolutionize fields such as healthcare, physics, and software development. Dr. Emily Hsu, a researcher in quantum optics, praised the model's capabilities. "The o1-preview has dramatically changed how we approach creating formulas necessary for our work. Its ability to generate and reason through high-level mathematical concepts is unparalleled," she noted.
OpenAI's o1-preview heralds a new era in artificial intelligence, promising not just answers but reasoned, thoughtful solutions to some of the most challenging problems faced by professionals today. As this technology evolves, it stands to not only augment human capabilities but also transform the landscape of problem-solving across various industries.
