Chinese AI company Deepseek has shocked the technology world with its AI tool. This is the reason for the massive dip in tech company stocks in the US market. Nasdaq has faced a 3 per cent dip, their worst performance in the last 2 years. Almost 01 trillion us dollars have been lost due to a sell-off sparked by DeepSeek. Deepseek claimed the top AI app on the Apple App Store within a few days by dethroning ChatGPT. DeepSeek-V3 and DeepSeek-R1 rival OpenAI’s cutting-edge models o1 and o3, as the Chinese lab achieved this feat only with a fraction of their investments.
Stats on DeepSeek
- Within a few days of launching it has crossed 10 million userbase worldwide.
- It has witnessed over 3 million downloads since the launch and 80% increase in the last 2 days.
- Deepseek has been used 3x more than its rival perplexity.
- It has been trained and developed using 5 million us dollars only which is quite smaller in front of its competitors like Chatgpt, Gemini, and meta ai.
- Deepseek has outclassed the GPT-4o and the Claude 3.5 sonnet in various benchmarks and surprised everyone.
- US Navy has banned the use of DeepSeek due to security and ethical concerns due to modal origin.
What is DeepSeek?
It’s an AI company located in Hangzhou, China founded by Mr. Liang Wenfeng. It has been reported that Liang has been working on DeepSeek since 2019 with his organization High Flyer AI. Liang and the shareholder have control over DeepSeek where whereas HighFlyer has a patent over the chipset technology used to train AI modals. DeepSeek is an open-source modal that allows anyone to use and build something which is not the case in Chatgpt. As a result, it has received acceptance from most developers around the globe.
Due to this high demand, cyber attackers were also attracted to DeepSeek which forced the company to limit the registrations. The company resolved issues relating to its application programming interface and users’ inability to log in to the website, according to its status page. The outages on Monday were the company’s longest in around 90 days and coincided with its sky-rocketing popularity.
ChatGPT vs DeepSeek AI vs Google Gemini

Feature | ChatGPT | DeepSeek | Google Gemini |
---|---|---|---|
Developer | OpenAI | DeepSeek | |
Model Version | Latest proprietary model | R1 (open-source) | Gemini 1.5 Pro |
Cost to Users | Subscription-based ($20/month) | Free | Subscription-based |
Performance | High performance across various tasks | Comparable to leading models, achieved with lower resource consumption | Advanced capabilities, including code execution and image generation |
Censorship | Provides uncensored information | Censors politically sensitive topics related to China (theguardian.com) | Provides detailed and uncensored information |
Integration | Standalone application | Standalone application | Seamless integration with Google services like Docs, Sheets, and Gmail (lifewire.com) |
Training Cost | High, estimated between $100 million to $1 billion | Significantly lower, approximately $5.6 million (thetimes.co.uk) | Not specified |
Open-Source | No | Yes | No |
Checkout – AI-Powered Job: Man Secures 50 Interviews After Applying to 1,000 Jobs While Asleep
What makes DeepSeek AI Modals Different?
With the limited resources and limited funding in front of AI powerhouse OpenAI, DeepSeek launched its DeepSeek-V3 model which has beaten the GPT-4o in various benchmarks. It is so powerful because of its architecture known as the Mixture of Experts(MOE). MOE is like group of experts working together to answer a query. The DeepSeek-V3 model is trained on 14.8 trillion tokens, which includes large, high-quality datasets that offer the model a greater understanding of language and task-specific capabilities.
It has an advanced technology to compete with the current models known as Multi-Head Latent Attention which cuts costs and increases efficiency.
Now they have launched the new modal DeepSeek R1 which has the about-thing known as test time computing. It costing this has been 95% more affordable than Chatgpt. It not only shows the result but also the chain of thoughts which shows it can think.
How DeepSeek is Cheaper
- Cost is reduced because of the uses of the NVIDIA H800 chipset an older chipset in comparison to the AI GPU NVIDIA H100 which is used by most of the big AI firms.
- DeepSeek engineers reportedly relied on low-level code optimizations to enhance memory usage.
- Chinese AI company has only trained the necessary parts of its model employing a technique called Auxiliary-Loss-Free Load Balancing.