DeepSeek quietly releases new version of AI

(Dan Tri) - Artificial intelligence startup DeepSeek has just launched an AI model with amazing power.

This new model not only challenges the dominance of American giants but also reshapes the entire game with open source (Illustrative image: DS).

This model has 685 billion parameters and is released under an open-source license. Initial test results show performance comparable to proprietary systems from OpenAI or Anthropic, and it opens up new approaches to developing and distributing advanced AI models.

Performance and cost

Immediately after its release on the Hugging Face platform, DeepSeek V3.1 achieved a 71.6% score in the Aider programming test. Technically, the model is capable of handling context with up to 128,000 tokens (equivalent to a 400-page book). The structure is presented as a "hybrid," integrating chat, inference, and programming functionalities.

AI researcher Andrew Christianson stated on social media platform X: "DeepSeek V3.1 achieved 71.6% on Aider, about 1% higher than Claude Opus 4, while costing significantly less."

According to calculations, the average cost to complete a programming task using V3.1 is around $1.01, while comparable systems can cost nearly $70.

Some community developers have also noted the presence of four “special tokens” in the architecture, enabling the model to integrate real-time web search and handle complex inference processes.

This difference translates into millions of dollars in savings for businesses that handle thousands of AI interactions every day.

Game-changing advantage

The release of V3.1 coincided with OpenAI and Anthropic announcing new models. By choosing an open-source approach, DeepSeek pursued a different strategy compared to many American companies that commercialize AI models through paid APIs.

Some argue that widespread availability of advanced capabilities can facilitate faster deployment by organizations and businesses, while also posing a challenge to business models based on technology monopolies.

The emergence of DeepSeek V3.1 shows that advanced AI is not just for large corporations, but smaller research groups can also participate.

Competition in this field is no longer limited to building the most powerful system, but also revolves around the ability to make technology accessible and affordable.

In this context, the question for technology companies is how to maintain commercial value when more and more open-source models with competitive performance and cost are emerging.

Source: https://dantri.com.vn/cong-nghe/deepseek-am-tham-ra-phien-ban-ai-moi-20250820121017977.htm