DeepSeek Revenue and Usage Statistics (2025)

David Curry

Updated: January 30, 2025

A new chatbot from Chinese AI lab DeepSeek sent shockwaves through the market in January, due to its ability to perform mathematics, coding and reasoning at a similar level to ChatGPT and other top-tier chatbots while costing far less and using fewer resources.

The chatbot launched globally on the 20th of January and has quickly drawn in a lot of users, with claims it matches ChatGPT in performance luring a lot of AI testers. The underlying model, which is in its third iteration, has been worked on by DeepSeek since November 2023.

DeepSeek is wholly funded by High-Flyer, a Chinese hedge-fund which utilizes AI for trading algorithms. Its founder, Liang Wenfeng, has made each iteration of DeepSeek open-source and available to view, modify and use for free. This has been instrumental in lowering the cost of AI models in China, with ByteDance, Tencent, Baidu and Alibaba cutting prices to compete.

Wenfang has not committed to any form of commercialization for DeepSeek, which may have kept it from being tied down by the Chinese regulators.

The latest iteration of DeepSeek used about 671 billion parameters and 14.8 trillion tokens, according to its post on transformers library Hugging Face. By comparison, OpenAI’s GPT 4 was reportedly trained on over one trillion parameters with 13 trillion tokens. Not huge differences, but in terms of cost, DeepSeek V3 cost $5.5 million, while GPT 4 cost over $100 million.

Part of this comes down to the smaller team and cost per engineer, but it’s also clear that DeepSeek was able to reach similar performance levels with far less processing power. It has less cloud capacity than OpenAI, and also does not have access to the latest Nvidia GPUs, such as the H100 and H200.

Reaching similar performance has drawn some to speculate that these advanced large language models do not need as much computing power as has been stated, which brings into question the hundreds of billions of dollars major tech companies are allocating to data center development and AI deployment. Nvidia, the biggest supplier of hardware for AI, saw its stock price drop by 18 percent in one day to a four month low.

It has also drawn into question the United States export controls on advanced chips to China, which was brought into law as a way to slow down the development of AI in China and ensure American companies continue to have a lead in AI.

We have collected data and statistics on DeepSeek. Read on below to find out more.

DeepSeek Key Statistics

  • DeepSeek third iteration cost $5.5 million to build, 1/18th the cost of OpenAI’s GPT 4
  • DeepSeek has been downloaded over 10 million times
  • DeepSeek is most popular in China, with 39% of iOS downloads coming from that country

DeepSeek Overview

Title 1Title 2
Launch date1 May 2023
HQHangzhou, China
PeopleLiang Wenfeng (founder, CEO)
Business typeSubsidiary
OwnerHigh-Flyer
InudstryAI

DeepSeek Users

DeepSeek had an average of 1.8 million daily active users since its launch in the middle of January.

DeepSeek Downloads

DeepSeek has been downloaded over 10 million times in the opening half a month.

DeepSeek downloads 2025 (mm)

DateDownloads (mm)
January 202510.7

Source: AppMagic

DeepSeek Downloads by Country

Close to 40% of DeepSeek downloads came from China, with the US accounting for 16% of total downloads.

DeepSeek downloads by country 2025 (%)

CountryUsers (%)
China39%
United States16%
India10%
South Korea7%
Russia6%
France4%
United Kingdom3%
Brazil3%
Germany3%

Source: AppMagic

DeepSeek vs ChatGPT

DeepSeek cost far less to build than GPT 4, even with similar tokens used.

DeepSeek V3 and OpenAI GPT 4 statistics

DeepSeek V3GPT 4
Parameters671 billion1 trillion
Tokens14.8 trillion13 trillion
Cost$5.5 million$100 million

Source: Company data

DeepSeek FAQ

Which country is DeepSeek based in?

DeepSeek is a Chinese based company, with company headquarters in Hangzhou.