OpenAI has announced the release of a new series of artificial intelligence models called“OpenAI o1“, known as“Strawberry“. This development is a breakthrough for AI, following the introduction of new models that simulate reasoning skills, at PhD levels in areas of study, such as mathematics, science and coding.
The launch of OpenAI o1 and its mini version
On September 12, OpenAI introduced this new generation of models with one focus: devoting more time to the thinking process to solve complex problems. Unlike previous models, OpenAI o1 has been trained to refine its reasoning, try different strategies and recognize errors, just as a person would. According to the company, this enables the model to tackle more difficult problems efficiently and accurately.
[post_relacionado]During OpenAI tests, the model demonstrated outstanding performance. For example, in the International Mathematical Olympiad, the new OpenAI o1 model scored 83% correct, compared to the 13% achieved by the previous previous GPT-4. In coding competitions such as Codeforces, the model achieved 89%, excelling in software development and complex mathematical problem solving.
OpenAI has also developed a faster and more affordable version called“OpenAI o1-mini“. This model, although “less powerful” than the full version, still offers excellent performance, especially in coding tasks, and is 80% more cost-effective, making it an attractive option for applications that do not require extensive knowledge of the world.
On the other hand, these new versions have achieved improvements in security. OpenAI has leveraged the reasoning capabilities of the new model to strengthen adherence to security guidelines, scoring better injailbreaking resistance tests. In one specific test, the o1-preview model significantly outperformed its GPT-4 predecessor, with a score of 84 compared to 22.
The construction of OpenAI o1, according to the engineers. Source: OpenAI
ChatGPT Plus users can access the o1 and o1-mini models, with weekly usage limits to maintain performance. Software developers with access to API level 5 can start working with both models for prototypes, although some features such as function calls and streaming are not yet available.
Follow us on social networks and don’t miss any of our publications!
YouTube LinkedIn Facebook Instagram X
Source and photo: OpenAI