Zuckerberg, fight with DeepSeek

Smart Things Author | Li Shuiqing Editor | Yunpeng Smart Things reported on August 6 that today, Meta launched its first programming Agent - Muse Code (beta version), as well as the new generation model Muse Spark 1.2 equipped with the Agent. As soon as the model was released, it sparked heated discussions overseas. The core focus was on the pricing strategy: cheaper than DeepSeek-V4-Flash. If users allow Meta to use their data to train models, they can enjoy a 5% discount. Interestingly, on the same day, DeepSeek issued an announcement announcing: "It plans to increase the overall pricing of API services in the near future, and the increase is expected to be larger." Other domestic head models also show signs of loosening the price war. For example, the price of the Zhipu GLM model will also increase significantly after the discount ends. When overseas giants OpenAI and Meta have entered the price war, domestic models such as DeepSeek seem to be out of step with the routine, and this global large-scale model melee has become more and more interesting. 1. Price reduction up to 75 times, Meta focuses on user data. Zuckerberg said in a tweet: "(This model) is positioned to be easy to get started and low-cost. (Users) can install it with just one line of code and start using the 'contributor package.'" The so-called "contributor package" means that Meta provides two API versions: one version allows data sharing, and the other version does not. The input price of the former is US$0.10/million Tokens, which is 12.5 times cheaper than the latter; the cache input price is US$0.002/million Tokens, which is 75 times cheaper; the output price is US$0.20/million Tokens, which is more than 21 times cheaper. Some netizens joked: "It's so funny, Muse Spark's input and output costs are lower than DeepSeek-V4-Flash, and it also has a uniquely optimized cache hit rate. They don't care about profits, they just want data." Previously, on July 31, DeepSeek-V4-Flash officially launched the API.