What Makes the "O1" Model Different from Other AI Models?

The "O1" model is one of the deep language models, meaning it operates on a similar foundation as other models. It can be accessed through the familiar ChatGPT interface (Shutterstock).

A new AI model has been unveiled by OpenAI under the name O1, a model that has been the subject of rumors for years, previously referred to by its code name Strawberry.

The company proudly claims that this model is capable of logical thinking and delivering more accurate and rational responses compared to any other AI model in the world, making it more precise in its answers.

But why does this upgrade deserve a standalone model, rather than a minor update to the usual ChatGPT model as with the previous 4.0 upgrade? What are the key differences between O1 and 4.0 ?

What Is Meant by Logical Thinking in "O1"?

The O1 model is a deep language model, meaning it fundamentally shares similar core technology with other models and can be accessed through the usual ChatGPT interface, though it requires a specific subscription.

While OpenAI has used several promotional phrases to describe this new model, such as its ability to think logically and verify the accuracy of responses before delivering them, it still closely resembles the standard ChatGPT model.

The company emphasized that logical thinking is the standout feature of the new model, setting it apart from other models, both from OpenAI and its competitors. While OpenAI hasn't fully disclosed the mechanics behind this logical thinking ability, it has outlined its basic framework.

The O1 model can break down complex questions into a series of smaller, simpler ones, answering each step-by-step while cross-referencing its findings with information from the internet before delivering a final response. While this method enhances the accuracy of the responses, it also takes more time.

This approach allows the model to tackle more complex questions, such as philosophical inquiries involving multiple aspects. The model breaks them down into smaller problems, addresses each, and then provides a comprehensive answer.

Traditional AI models work by receiving the main question, which we'll call "input," and then searching the internet for answers (Getty).

What Differentiates "O1" from ChatGPT?

At first glance, OpenAI’s new model seems similar to other AI models but with better accuracy. However, the slight change in how the model answers questions—what OpenAI calls "logical thinking"—makes "O1" significantly different from all other models.

Though OpenAI hasn’t fully revealed the details of its logical thinking mechanism to protect its proprietary advantage, it has provided a general overview. To understand the distinction between this new process and traditional AI models, one must first consider how standard AI works.

Traditional AI models process a primary question (input), search the internet for relevant information, and match keywords from the question with content online to produce an answer (output). OpenAI has typically used this method in pricing models, but O1 introduces a different approach.

O1 employs a process known as Chain of Thoughts, which mimics how humans solve problems by connecting various ideas to form a solution. This process can be likened to solving a math puzzle with multiple steps, such as figuring out: Mohammed is 10 years old, Ali is 4 years younger, so how old is Ahmed if he is two years older than Ali?

Answering such puzzles requires a series of calculations that are logically connected, and this is how O1 operates. Although this method results in more accurate responses, it also takes longer than traditional models, which provide answers almost instantly.

This process poses a new challenge for OpenAI, as it involves hidden steps that are not visible to the user. Therefore, the pricing for O1 cannot rely on the usual input/output tokens. OpenAI introduced a new pricing concept known as Reasoning Tokens, which account for the unseen background operations. Due to these additional processes, the cost of using O1 is four times higher than that of standard ChatGPT.

The "O1" model can break down complex questions and problems into simpler, smaller questions. It then answers these smaller questions and compares its responses with information from the internet (AFP).

Unique Use Cases

The logical thinking mechanism in O1 opens the door to a wide variety of use cases, pushing AI into more professional applications, such as medical, mathematical, or scientific problem-solving.

The Chain of Thoughts process allows the model to answer complex, multi-layered questions by breaking them down into smaller, manageable tasks. This process ensures higher accuracy in most cases.

Although OpenAI integrated this mechanism into its new model, it's not entirely new to AI. Users have been training AI models to use similar methods for better results, though it requires more time and effort than traditional techniques.

The pricing of O1 also shifts its use toward more professional applications, as its cost is four times higher than the usual ChatGPT model, making it the most expensive AI model currently available. It’s worth noting that the model is still in its experimental phase, and future versions are expected to be more efficient.