DeepSeek for iOS has overtaken OpenAI’s ChatGPT to top the App Store’s “Top free apps” chart in the US. The eponymous Chinese company released the open-source DeepSeek-R1 artificial intelligence (AI) model last week, which outperforms OpenAI’s o1 AI model in several benchmarks. This release has made the company and its AI chatbot the talk of the town, with several Silicon Valley tech leaders reacting to its sudden rise. Notably, the DeepSeek app is entirely free to use, with no subscription tiers announced so far.
DeepSeek Overtakes ChatGPT
Ever since OpenAI launched the ChatGPT for iOS app in May 2023, it has consistently been among the top free applications on the App Store and has remained the highest-ranked AI app on the platform. However, this throne was taken by the DeepSeek app recently as it reached the top of the chart. The rise is being attributed to the recent launch of the DeepSeek-R1 AI model.
Notably, little is known about the Hangzhou-based AI firm which was founded in 2023 and has released several open-source large language models (LLMs). While US-based tech companies have also released open-source models, with Meta being a notable mention in the community, none comes close to the capabilities and scale of what the DeepSeek models offer. The company also claimed to have built the model with a cost of $6 million (roughly Rs. 51.8 crore), significantly cheaper than AI models of this scale.
Several Silicon Valley-based tech leaders have reacted to the sudden rise of these models. Venture capitalist Marc Andreessen called the R1 AI model “one of the most amazing and impressive breakthroughs I’ve ever seen,” while Perplexity AI Co-Founder and CEO Aravind Srinivas congratulated the Chinese firm for becoming the first AI app to beat ChatGPT.
In a LinkedIn post, Yann LeCun, one of the godfathers of AI and the current Vice President and Chief AI Scientist at Meta, said “Open source models are surpassing proprietary ones.” Holger Zschäpitz, a senior financial reporter at German free-to-air television news channel Welt, argued DeepSeek “could represent a big threat to US equity markets.”
DeepSeek Is Not the Only One
While DeepSeek has broken the trend of companies open-sourcing smaller and less capable AI models while keeping the frontier models under a proprietary paywall, it is not the only one. Another Chinese firm, Kimi AI, announced the release of the Kimi k1.5 AI model. The company claims it is an “o1-level multimodal LLM that outperforms GPT-4o and Claude Sonnet 3.5 on several benchmarks”.
Notably, Kimi AI has made the web version of its chatbot completely free to use with unlimited usage, similar to DeepSeek. The AI model can also perform real-time web searches, analyse up to 50 files across a diverse range of formats, and has image understanding capabilities as well. While its technical report is available on GitHub, it is currently not available in open-source.
