Neither Chatgpt nor Gemini: The Chinese the assistant who occupies the first position in the App Store

The Depseek wizard arrived at App Store On January 11 and since then, he has climbed to occupy the first position in the Apple store in the United States, ahead of Chatgpt, from OpenNai; A milestone for an application that has just reached the market and that competes with the most popular ‘chatbot’ for more than two years.

Deepseek is an artificial intelligence assistant (AI) that the Chinese firm of the same name launched on January 11 at the App Store, where it is offered under a Free use modality. In addition to answering questions in a conversational format, it can Navigate the web to offer updated responses, summarize text documents quickly and use reasoning to solve complex problems.

At its base is the Deepseek V3 model, which was launched in December. This has been trained with 671,000 million parameters With Mixture of Experts (MOE) architecture, which divides an AI model into small neural networks to act separately as if they were different experts.

It also has 37,000 million activated parameters for each token, as their person responsible explains in the Github repository. And those responsible have resorted to the multicabezal latent care mechanism (MLA) to “achieve efficient inference and profitable training.”

The technology company has used 14.8 billion “diverse and high quality” tokens along with a supervised adjustment and phase reinforcement learning. They also ensure that each Token has required for their training 3.7 days with 2,048 GPU NVIDIA H800, which makes the Total training reach 2,788 million hours of GPU and a total cost of 5,576 million dollars.

As they claim, the “exceeds other open source models and achieves a performance comparable to that of the main closed code models.” Thus, in the evaluation of language understanding (MMLU PRO) in a variety of tasks, Deepseek v3 reaches a 75.9 score at 78.0 of Claude 3.5 Sonnet, the 72.6 of GPT 4 and the 73.3 call 3.1 405b.

In the evaluation of the ability to answer complex questions of Postgraduate level (Gpaq diamond), Deepseek v3 has obtained a 59.1 score, Below Claude 3.5 Sonnet (65.0), but above GPT 4 (49.9), Qwen 2.5 of 72b (49.0) and calls 3.1 405b (51.1).

It also remains in second position in the analysis of the ability to solve software problems of the real world (SWE), in which it reaches a score of 42.0, compared to the 50.8 of Claude 3.5 Sonnet, followed by GPT 4 (38.8), calls 3.1 405b (24.5) and Qwen 2.5 of 72b (23.8).

On the other hand, on the challenge resolution tests (Math 500), where he obtains 90.2, while Claude 3.5 Sonnet reaches 78.9; Qwen 2.5 of 72b, 80.0; GPT 4, 74.6, and call 3.1 405B, 73.8 points. He also does so in solving mathematical problems with Aime 2024, with a score of 39.2, followed by Qwen 2.5 of 72b and calls 3.1 405b (both 23.3), Claude 3.5 Sonnet (16.0) and GPT 4 (9.3).

In programming, in the Codeforces test, Depseek V3 reaches 51.6 points, Qwen 2.5 of 72b, 24.8; Call 3.1 405B, 25.3 points; GPT 4O23.6; and Claude 3.5 Sonnet, a score of 20.3.

New reasoning models

Last week, the Chinese company presented a new family of reasoning models, Depseek-R1-Zero and Deepseek-R1, the first one trained through large-scale reinforcement learning without supervised fine adjustment as a preliminary step, as explained In the research text published in Arxiv.org.

In the second, on the other hand, training in multiple phases and cold starting data before the learning of reinforcement to overcome the problems of legibility and mixture of languages ​​have also been used. Thanks to this, its developers say that it reaches a performance in reasoning tasks comparable to OpenAi O1.

“Our goal is to explore the potential of the LLM [modelos de lenguaje de gran tamaño] To develop reasoning capabilities without any supervised data, focusing on their self -evolution through a pure reinforcement learning process, ”they explain.