Two years ago, when DeepSeek, which had not yet entered the market, started a price war with the attitude of a “price butcher,” probably few people would have imagined that the big domestic model could close the gap with the top foreign models to less than 3 months in just 2 years, and at the same time have the right to increase prices. Recently, China's big model has made new breakthroughs one after another. On July 17, Dark Side of the Moon released the Kimi K3 model. The model has 2.8 trillion parameters and a context window of 1 million words, and is the open source weight model with the largest number of parameters in the world. On July 19, Ali Qianwen launched the Qwen3.8-Max preview model and announced that Qwen3.8 will soon be released and open sourced. The model's parameter scale reached 2.4 trillion yuan, “probably the most powerful model other than Fable 5.” The trillion-level parameter scale and performance comparable to top closed source models give domestic AI the incentive to raise prices. The application interface call cost of the K3 model has increased dramatically compared to the previous K2.6 price, and it is currently the most expensive domestic model. According to industry insiders, this is an important sign that the big domestic model is changing from a cost-effective alternative to a technical parallel runner. However, on this crowded racetrack, the distance between players is constantly shortening, which can almost be described as “close physical combat.” When technological leadership can be quickly disrupted weeks or even days later, how can big model companies prove their uniqueness? This not only determines how high the market values it, but also determines who can have the last laugh in this “leftovers are king” knockout game.

Two years ago, when DeepSeek, which had not yet entered the market, started a price war with the attitude of a “price butcher,” probably few people would have imagined that the big domestic model could close the gap with the top foreign models to less than 3 months in just 2 years, and at the same time have the right to increase prices. Recently, China's big model has made new breakthroughs one after another. On July 17, Dark Side of the Moon released the Kimi K3 model. The model has 2.8 trillion parameters and a context window of 1 million words, and is the open source weight model with the largest number of parameters in the world. On July 19, Ali Qianwen launched the Qwen3.8-Max preview model and announced that Qwen3.8 will soon be released and open sourced. The model's parameter scale reached 2.4 trillion yuan, “probably the most powerful model other than Fable 5.” The trillion-level parameter scale and performance comparable to top closed source models give domestic AI the incentive to raise prices. The application interface call cost of the K3 model has increased dramatically compared to the previous K2.6 price, and it is currently the most expensive domestic model. According to industry insiders, this is an important sign that the big domestic model is changing from a cost-effective alternative to a technical parallel runner. However, on this crowded racetrack, the distance between players is constantly shortening, which can almost be described as “close physical combat.” When technological leadership can be quickly disrupted weeks or even days later, how can big model companies prove their uniqueness? This not only determines how high the market values it, but also determines who can have the last laugh in this “leftovers are king” knockout game.