The OpenAI model doesn’t show its thinking, which gives open source an advantage

OpenAI has heralded a new paradigm of thinking in large language models (LLMs) by Model o1which recently received a major upgrade. However, although OpenAI has a strong lead in inference models, it may be losing some ground to them Open source competitors That appears quickly.

Models like o1, sometimes referred to as large reasoning models (LRMs), use additional inference time computation cycles to “think” further, review their answers, and correct their answers. This enables them to solve the complex inference problems faced by classical LLM students and makes them particularly useful for tasks such as programming, mathematics and data analysis.

However, in recent days, developers have shown mixed reactions to o1, especially after the updated version. Some posted examples of o1 accomplishing amazing tasks while others did just that Express frustration On confusing form responses. Developers have encountered all kinds of problems ranging from making illogical changes to code or ignoring instructions.

Confidentiality about details o1

Part of the confusion is due to OpenAI’s secrecy and refusal to show details of how o1 works. The secret sauce behind the success of LRMs is the additional codes that the model generates as it reaches the final response, which are referred to as the model’s “thoughts” or “logic chain.” For example, if you classically ask an LLM to generate code for a task, it will generate the code immediately. In contrast, LRM will generate inference tokens that examine the problem, plan the code structure, and generate multiple solutions before submitting the final answer.

o1 hides the thinking process and shows only the final response with a message displaying how long the model has been thinking and perhaps a high overview of the thinking process. This is partly due to avoiding response congestion and providing a smoother user experience. But more importantly, OpenAI considers the inference chain a trade secret and wants to make it difficult for competitors to replicate o1’s capabilities.

The costs of training new models continue to grow, and profit margins are not keeping pace, prompting some AI labs to become more underground in order to expand their advances. Even Apollo research, which… Red team of the modelis not granted access to its logic chain.

This lack of transparency has led users to make all kinds of speculation, including accusing OpenAI of weakening the model to reduce inference costs.

Completely transparent open source models

On the other hand, open source alternatives like Alibaba Quinn with questions and marco-o1 View the complete logic chain of their models. Another alternative is Deep Sec R1which is not open source but still exposes logical codes. Visibility of the chain of reasoning allows developers to troubleshoot prompts and find ways to improve form responses by adding additional instructions or examples in context.

Visibility into the inference process is especially important when you want to incorporate model responses into applications and tools that expect consistent results. Moreover, controlling the underlying model is important in enterprise applications. Special models and the scaffolding that supports them, such as protections and filters that test their inputs and outputs, are constantly changing. Although this may improve overall performance, it may break many prompts and applications built on top of them. In contrast, open source models give full control of the model to the developer, which can be a more powerful option for enterprise applications, where performance on very specific tasks is more important than general skills.

The QwQ and R1 are still in preview versions and the o1 takes the lead in terms of accuracy and ease of use. For many uses, such as providing custom generic claims and one-time requests, o1 can still be a better choice than open source alternatives.

But the open source community is rushing to catch up with proprietary models and we can expect more models to emerge in the coming months. It can turn out to be a convenient alternative where visibility and control are crucial.