OpenAI presented a new ChatGPT o1 model: a revolution in the "thinking" of artificial intelligence

OpenAI - ChatGPT o1 is the company's first model capable of logical thinking and not just simple word production. It is intended for solving more complex problems, such as multi-step questions, demanding mathematical tasks and writing code. O1 uses a technique called “chain of thought”, which means that AI solves a problem step by step, just like humans do. With a new training method that includes optimization algorithms and reinforcement learning, the model can better analyze errors and understand context, resulting in greater accuracy in answers.

ChatGPT o1 – High performance for a higher price

O1 is designed for more demanding tasks such as research in science, coding and solving mathematical puzzles at the PhD level. Compared to GPT-4o, which was previously the flagship of OpenAI's models, o1 performs better on specific tasks – for example, it scored 83 % correct answers on the Math Olympiad tests, while GPT-4o barely managed 13 %​.

Of course, all this knowledge does not come cheap. O1-preview costs $15 per million input text characters and as much as $60 per million output, which is more than three times more expensive than GPT-4o. Although the o1-mini, its smaller version, will be more affordable, companies will still face high costs if they want to use the full potential of this system.

The first step towards the future of artificial intelligence?

What does this mean for the future of AI? OpenAI is clearly moving towards agents – autonomous systems that will be able to make decisions and perform tasks without human supervision. Although o1 is not yet at this level, it represents an important step forward. This could mean revolutionary changes in science, medicine and engineering, where artificial intelligence will be able to solve complex problems without the help of humans.

Of course, we must not forget that this is just the beginning. OpenAI admits that the model still hallucinates – meaning it sometimes produces wrong answers – and that it lacks several features such as web browsing or image processing. But the team promises further upgrades and development of models that will be even more useful in everyday applications.

Conclusion

With the release of the o1 model, OpenAI took a big step towards more advanced artificial intelligence capabilities, but the road to full automation and "human" intelligence is still long. For now, o1 can be used by ChatGPT Plus and Team subscribers, while access for free users will come later. For anyone who can afford the price, however, this is certainly one of the most impressive AI models to date.