Deepseek Artificial Intelligence became the most downloaded US AppStore app

Less than a month ago, the Chinese artificial intelligence startup Deepseek released its chat app. This chat, though free, has been able to compete with the powerful models of Openai. Now the iOS version of its iOS has won first place.

According to Reuters, the Deepseek Artificial Intelligence app was able to cross the ChatGpt and become the most downloaded free app in Apple’s App Store. These chats use the Deepseek-V3 model, which, according to the makers, “ranks first among the open source models and is the most advanced competitor of the world’s most advanced packages.” This app has gained popularity among users since January.

Deepseek app in the first place AppStore Apple

Artificial intelligence models such as ChatGpt and Deepseek need advanced chips such as the Nvidia H100 for training, but the US government has expanded the export bans to China since 2021. Due to the sanctions, China cannot have the H100, in addition to importing the H800, which has lower data transfer rates. Researchers at Deepseek also say the Nvidia H800 chips used to train the Deepseek-V3 model and cost less than $ 6 million.

Despite the hardware limitations of Deepseek Startup, the app has been able to cross ChatGpt at AppStore. This has worried the big US corporations and has led to a decline in future trading trading index. The popularity of Chinese artificial intelligence has caused the efficiency of US sanctions to be questioned. On the other hand, with the release of DeepSic’s pioneering news, Chinese technology companies related to Deepseek, such as IFlytek, increased.

The strange thing is that Deepseek, despite fewer financial resources and hardware, also unveiled a few days before the R1 open -ended model, which in some benchmarks defeated the Openai O1 reasoning model; It also costs 95 % cheaper than the Openai model.

Now we have to see what American companies are approaching to counter this Chinese startup and will make a change in pricing their models.