The model operates with 685 billion parameters and, according to DeepSeek, stands out with faster processing capabilities and enhanced “smart agent” functions. In other words, it can complete tasks more quickly and accurately, while better understanding complex instructions.
The V3.1 model was released directly on the Hugging Face platform without any major marketing or launch event. Early tests indicate that this open-source model can compete with proprietary AI systems from major companies like OpenAI and Anthropic.
The new model is also offered at accessible pricing for users. The company has announced new API usage rates: input API costs $0.07 per 1 million tokens with caching, $0.56 per 1 million tokens without caching, and output API costs $1.68 per 1 million tokens. This makes it a convenient option for developers and anyone looking to integrate AI into web or mobile applications.
Experts believe that the V3.1 model will further strengthen DeepSeek’s position in the market, especially due to its technical performance and ease of use, allowing the company to compete more effectively with rivals.
This step also highlights the importance of open-source technologies in AI and opens new opportunities for developers.
