Photo credit: www.techradar.com
Similarweb data shows DeepSeek’s usage doubled to 12 million users globally in just two daysThe launch of DeepSeek’s R1 model has incited concern among Western AI industry leaders about its financial implicationsWhile ChatGPT enjoys higher total page views, DeepSeek is gaining rapidly
In a remarkable turn of events over a two-week period, the open-source, MIT-licensed Chinese large language model (LLM) DeepSeek has captivated the AI community, leading to notable fluctuations in the stock of Western AI giant Nvidia and inducing OpenAI CEO Sam Altman to accuse DeepSeek’s developers of potentially leveraging their models for training.
Western AI executives are understandably on alert, especially with recent statistics from Similarweb revealing significant growth for DeepSeek’s centralized web and mobile applications. The essence of open-source technology allows users to run various models locally, which may limit the visibility of some traffic metrics for these setups.
Comparing ChatGPT and DeepSeek
The pivotal moment came at the end of January, marking the release of DeepSeek’s advanced ‘R1’ model. Between January 24 and January 26, 2025, global daily visits to DeepSeek surged from 6.2 million to 12.4 million.
Prior to this spike, there was a notable decrease in page views during the week leading up to the model’s launch, with a drop of around 900,000 visits from January 15 to 18. However, from January 19 onward, just before the launch, DeepSeek enjoyed consistent growth, culminating in the impressive two-day increase.
In contrast, ChatGPT has seen sharp declines in traffic coinciding with the launch of the R1 model. Data indicates that during the same timeframe, ChatGPT experienced a drop of 43.1 million views between January 15 and 18, followed by a significant decline of 41.3 million views from January 23 to 25.
These fluctuations might be part of the natural ebb and flow of web traffic; however, it’s noteworthy that, despite the downturn, ChatGPT’s losses still vastly exceed DeepSeek’s total traffic, suggesting that it remains a dominant force. On January 23, just three days after DeepSeek’s R1 release, ChatGPT reached an impressive 140.6 million views.
The primary concern revolves around growth trends. ChatGPT appears to be plateauing, recording an average of 126.9 million views during the week of DeepSeek’s R1 release, with only sporadic spikes around 140 million views—numbers that do not reflect growth compared to previous months, as indicated by Similarweb data.
DeepSeek, while averaging 7.45 million views during this timeframe, poses a growing threat to its rivals, raising questions about the sustainability of its growth and whether it can establish itself as a formidable competitor in the long run. The enthusiasm surrounding a locally-hosted model may foster a competitive spirit in the field, but ChatGPT’s established presence complicates DeepSeek’s ambitions.
Understanding DeepSeek
For those unfamiliar, DeepSeek is a Chinese enterprise that has developed its R1 model to be “on par with ChatGPT” and other leading models, achieving this for a mere $6 million (£4.8 million) in training expenses—significantly less than the $78 million invested by OpenAI for their latest GPT-4. This achievement is notable amidst ongoing embargoes of advanced GPU technology to China.
Founded in May 2023 by Liang Wenfeng, DeepSeek’s early success is partly attributed to his strategic acquisition of Nvidia GPUs in 2021 out of “curiosity,” prior to the imposition of export restrictions on semiconductors by the U.S. in October 2022.
DeepSeek R1 is freely available to small and medium-sized businesses on Github, and being MIT-licensed offers a significant advantage for the implementation of low-cost AI chat solutions. What sets its model apart is the capability of running locally and effectively on modest hardware, unlike its competitors.
While concerns surrounding data leaks have surfaced, particularly regarding its online and Android applications, running the model locally provides users a means to potentially mitigate those risks.
You might also like
Source
www.techradar.com