The OpenAI o1 model hides the thought process, giving an advantage to open-source solutions

OpenAI introduced a new approach to problem solving with large language models (LLMs) in its o1 model, which recently received a major update. However, despite OpenAI’s leadership in creating models with powerful reasoning abilities, open source analogs are beginning to rapidly gain popularity, offering their own advantages.

Features of the o1 model

Models like o1, often referred to as large-scale reasoning models (LRMs), use additional computational cycles during query execution. This allows them to “think”, analyze tasks, check their answers, and correct errors. Because of this, they successfully solve complex problems such as programming, math, and data analysis that traditional LLMs have difficulty with.

These models can be used to solve complex problems such as programming, math, and data analysis.

Developers, however, have expressed mixed feelings about the updated version of o1. Some praise its incredible abilities, while others complain about illogical answers, ignoring instructions, or strange changes in the code.

A number of developers have complained about the new version.

Secrecy of OpenAI

A major cause of dissatisfaction is the secrecy of OpenAI. The o1 model hides the thought process, giving users only the final answer and the total running time of the model. This is done to simplify the user experience and protect trade secrets. The reasoning chain – the process of generating intermediate “thoughts” that help the model arrive at a final decision – is seen by OpenAI as a key competitive asset.

Even the teams involved in testing the model did not have access to this data, prompting speculation that OpenAI may have reduced the quality of the model’s performance to save computational costs.

Even the teams involved in testing the model did not have access to this data, prompting speculation that OpenAI may have reduced the quality of the model to save computational costs.

Transparency in open-source models

Open-source alternatives, such as Qwen with Questions from Alibaba or DeepSeek R1, provide full access to the reasoning chain. This allows developers to analyze intermediate model steps, optimize queries, and improve results. Transparency is especially important for enterprise applications where stability and predictability of model performance are critical.

.

Private models provide full control over their internal processes, making them more flexible for specific tasks. Unlike private solutions, their customizations can be tailored to meet specific business needs without the risk of disruption from sudden changes.

Public models provide full control over their internal processes, making them more flexible for specific tasks.

Competition between private and open solutions

.

The o1 model remains the leader in terms of accuracy and usability for common tasks such as one-off queries or generating answers to ad-hoc questions. However, open-source solutions are quickly catching up, and new competitive models are expected to emerge in the coming months that will offer developers more control and flexibility.

The o1 model remains the leader in accuracy and usability for common tasks, such as ad hoc queries or generating answers to ad-hoc questions.