The first OpenAI model with reasoning capability called o1 was released + video

OpenAI, a new artificial intelligence model named o1 has introduced the first model of this startup from a series of models with the abilityreasoning» to answer more complex questions. Also, a smaller and cheaper version of it called o1-mini has also been published. In the rumors about this startup, this model was mentioned with the name “Strawberry”.

o1 is an important step for OpenAI towards the goal of producing human-like artificial intelligence models. This model performs better than previous models in writing code and solving multi-step problems, but it is more expensive and slower than GPT-4o. OpenAI also refers to o1 as a “preview” version to emphasize its nascent nature.

The video below shows o1’s ability to generate code for a video game from a simple text command:

The cost and time to access the o1 model

ChatGPT Plus and Team users have access to o1 and o1-mini now, and Enterprise and Edu users will get access early next week. OpenAI also plans to make o1-mini available to all free ChatGPT users, but has yet to announce when.

For developers, this model is offered at a fairly high price, and its API is available for every 1 million input tokens. $15 And for every 1 million outgoing tokens 60 dollars It costs money. For comparison, GPT-4o costs $5 per 1 million input tokens and $15 per 1 million output tokens.

Also, the method of training this model is different from previous OpenAI models, although this startup did not provide a detailed explanation about its details. Jerry Turk, head of OpenAI research, says that o1 was trained using a completely new optimized algorithm and a new training data set designed specifically for it.

In fact, for o1, a technique called “reinforcement learning” has been used, which makes the model solve problems by itself with rewards and penalties. It then uses a “chain of thought” to process the questions, much like humans process problems in a step-by-step review.

With this new training method, OpenAI says the model performs more accurately and has fewer illusions.

The following video shows the reasoning capability of this model:

According to OpenAI’s description, the most important thing that differentiates this new model from GPT-4o is its ability to solve complex problems, such as coding and math, which performs much better than previous models, as well as explaining its reasoning. However, o1 is not as capable as GPT-4o in many areas. o1 doesn’t do well in real-world knowledge, nor does it have the ability to browse the web or process files and images.