The arrival of ChatGPT in 2022 marked a significant shift in artificial intelligence, but it’s not the only player making waves. In early 2025, a lesser-known Chinese company named DeepSeek burst onto the AI scene, challenging established giants. If you’re curious about DeepSeek AI and its implications, you’re in the right place. Let’s dive deep into this emerging technology.
What is DeepSeek AI?
DeepSeek AI is an advanced AI chatbot launched by Hangzhou DeepSeek Artificial Intelligence Basic Technology Research Co., Ltd., based in Hangzhou, China. Under the leadership of CEO Liang Wenfeng, this startup emerged in July 2023 and has quickly positioned itself as a competitor to firms like OpenAI and Anthropic. In January 2025, the company gained significant traction with the release of its R1 reasoning model, which nearly matched OpenAI’s performance and soared to the top of the US App Store.
What sets DeepSeek apart is its cost efficiency. The company trained its R1 model for just $6 million, a fraction of the estimated $100 million required for similar models in the West, triggering a notable 18% decline in Nvidia’s stock price. Furthermore, DeepSeek has been transparent about its methodologies, providing the model weights for free to promote open-source development in AI.
While OpenAI charged users $20 to access its o1 reasoning model, DeepSeek offered R1 for free. This strategic move pressured other major players to reconsider their pricing structures and made powerful AI tools more accessible worldwide.
AI Models Released by DeepSeek
DeepSeek has been busy building its portfolio. Its first model, DeepSeek Coder, was designed for coding tasks, followed by the DeepSeek LLM and DeepSeek-Math. In just half a year, the company launched the more advanced DeepSeek-V2 and DeepSeek-Coder V2. By September, DeepSeek-V2.5 had made its debut, and in November, the first reasoning model, DeepSeek-R1-Lite, was released, alongside the DeepSeek-V3 base model in December.
Fast forward to January 2025, and the capable DeepSeek-R1 reasoning model was publicly available along with an app for Android and iOS. DeepSeek continues to refine its offerings, with the latest versions, DeepSeek-R1-0528 and DeepSeek-V3.2, released in September. Excitingly, rumors suggest a major new model may be on the horizon.
How to Use DeepSeek AI?
You can easily engage with DeepSeek AI for free by visiting chat.deepseek.com. Alternatively, download the DeepSeek app for your mobile device on Android or iOS. If you prefer, you can run distilled versions of DeepSeek R1 locally on your PC or mobile device.
What is Happening with DeepSeek Now?
Recently, DeepSeek has not launched any major models. The Chinese government has encouraged the company to use domestically produced AI chips like the Huawei Ascend for future developments. According to a report by the Financial Times, this request has led to delays in the release of the R2 model. Although DeepSeek aimed to adopt the Ascend chips, they reverted to Nvidia GPUs for training due to technical hurdles.
Currently, DeepSeek reportedly utilizes Ascend chips for inference while using Nvidia GPUs for training. As one of the most promising AI startups to emerge from China, DeepSeek is a company to watch, alongside other tech giants like Alibaba.
Curious to learn more about cutting-edge technologies and developments? Explore more content at Moyens I/O for the latest updates and insights.