o1 preview and o1 mini, the real changes?

OpenAI continues to push the boundaries of AI with the launch of its brand-new “Strawberry” model on September 12.

This model was designed to address complex challenges in areas such as science, programming, and math.

For now, only subscribers have access to this preview version, but many improvements are already under development.

‍

‍

How does it work?

What sets these new models apart is their ability to think longer before providing an answer. A bit like a human being who takes the time to analyze a problem before reacting. Thanks to this, they manage to refine their thinking processes, test different strategies and identify their mistakes. According to OpenAI, the next update of the model even achieved results comparable to those of doctoral students on complex tasks in physics, chemistry, and biology.

In mathematics and programming, the performances are impressive. For example, during a test for the International Mathematical Olympiad, the gpt-4O version solved only 13% of the problems, while the o1 model solved 83%! However, some practical features, such as accessing the Internet or downloading files, are not yet available.

But for tasks that require complex reasoning, this new model represents a major breakthrough for Generative AI.

‍

‍

For whom?

At the moment the o1 model is particularly useful for

Health researchers, to annotate cell sequencing data.
Physicists, to generate complex mathematical formulas in fields such as quantum optics.
The developers, regardless of their domain, to create and execute multi-step workflows.

‍

OpenAI O1-Mini

The o1 series is already recognized for its efficiency in code generation and debugging. To offer a faster and more affordable alternative to developers, OpenAI launched the model O1-Mini. This model, while smaller, is just as powerful for coding tasks.

It is especially 80% cheaper than the o1-preview version, making it ideal for applications that require reasoning but not in-depth knowledge.

‍

How to use it?

ChatGPT Plus and Team subscribers can now access the o1-preview and o1-mini models directly in the ChatGPT template selector.

However, there are weekly limits of 30 messages for o1-preview and 50 for o1-mini.

‍

‍

Eligible developers can already start experiment with these models via the API, with a limit of 20 requests per minute.

Although some features are still missing, OpenAI plans toexpand these capabilities very soon.

OpenAI also plans to make o1-mini available for free to all ChatGPT Free users.

‍

o1-preview & mini already in our courses?

At AI Sisters, we're constantly looking to improve our courses to keep them at the forefront of the latest innovations in generative AI.

That is why we are pleased to announce that as of today, we are integrating the o1 Strawberry model into our training courses! 🥳

Our aim? Accompany you at best so that you can not only understand, but above all master this new tool, and use it in your daily life to save precious time but also even more productivity !

‍