On September 24, two informed sources revealed that DeepSeek's current annual revenue run rate has reached $1 billion. Just a few months ago, this figure was below $500 million. The revenue growth is attributed to a recent increase in API pricing and the continued popularity of its large models in the market. DeepSeek CEO Liang Wenfeng disclosed the latest revenue data during a recent investor meeting. The company is advancing its second round of financing and preparing for an IPO on the Shanghai Stock Exchange, with this revenue performance expected to boost investor confidence. DeepSeek aims to complete this round of financing by the end of October, targeting to raise 50 billion yuan (approximately $7.5 billion) with a valuation goal of 500 billion yuan. According to insiders, part of DeepSeek's revenue growth comes from last month's price adjustment, which raised model invocation prices to 2.3 to 4.5 times the original rates. Liang Wenfeng stated at the meeting that the price adjustment has not led to customer loss, and user demand remains strong. Insiders reported that Liang Wenfeng informed investors that DeepSeek continues to allocate most of its resources to the development of new models, with over 70% of computing power dedicated to model training, leaving less than 30% for inference operations of existing models. Liang Wenfeng also indicated to investors that internal testing shows the company's lightweight models can run stably on gaming graphics cards, capable of handling the vast majority of users' daily tasks. (Sina Finance)