What is DeepSeek AI? Why is everyone talking about it this week? If these are questions you’ve been looking for answers to, you’re not alone. Interest in a new Chinese AI service is sweeping across the country, and its impact has been so substantial that Nvidia’s stock slumped almost instantaneously upon its release.
The DeepSeek AI and Nvidia connection will require further explanation, but what you need to know first is what the AI tool is and what it can do for you. If it sparks your interest, we’ll dive into how the DeepSeek AI vs OpenAI battle is taking shape and how to download the DeepSeek AI as well.
Image: Pexels
What is DeepSeek AI?
By now, most of us have used artificial intelligence to some degree, whether to play around with the ChatGPT app or use the handy tools that have been made available by smartphone makers directly on our devices. DeepSeek AI is a similar tool that Liang Wenfeng introduced through a company he founded in 2023 and headquartered in Hangzhou, Zhejiang.
Last month, DeepSeek released a new AI model that first grabbed the attention of experts on the field and kicked off insider discussion of its capabilities. New-found fame only arrived at their doorsteps this month, when the organization published a follow-up research paper that developed into the next version of their DeepSeek model, which is what we have access to now.
DeepSeek released the latest version of its Artificial General Intelligence (AGI) tool, DeepSeek-R1, which is based on the DeepSeek-V3 model. The V3 model showcases 671 billion parameters and was trained on a dataset of 14.8 trillion tokes within 55 days. The model was also significantly inexpensive to develop compared to others, the costs adding up to $5.58 million.
The benchmark tests showed the LLM model outperformed tests like Llama 3.1 and Qwen 2.5, coming up close to the capabilities of GPT-4o.
DeepSeek AI vs OpenAI, Which is Better?
The DeepSeek AI presents us with similar conversational capabilities we’ve seen from ChatGPT. The AI is able to reason out responses and communicate them in a way that is easy for the users to understand.
Both are free to use, however, ChatGPT’s more advanced features are locked behind paywalls. Even with DeepSeek, there are advanced pay-as-you-go features for more professional use cases if you need more out of the AI. Still, DeepSeek is more affordable to both produce and use.
The open-source nature of the framework is a key feature of the DeepSeek tool, giving users room to modify and build with it, and in this area as well, it comes out on top. ChatGPT still remains superior just as a result of the vast array of features it has, such as image generation or vision capablities, however, this is only the nascent stage of accessing DeepSeek’s capabilities.
Chiming in on the DeepSeek AI vs OpenAI debate, Sam Altman, CEO of the latter platform, agreed that it was good to see more competition in the AI industry. He also added that ChatGPT would continue to build on features that would set it apart.
deepseek’s r1 is an impressive model, particularly around what they’re able to deliver for the price.
we will obviously deliver much better models and also it’s legit invigorating to have a new competitor! we will pull up some releases.
— Sam Altman (@sama) January 28, 2025
What Has the Public Reception to DeepSeek Been Like?
The AI bot impressed experts in the field at first, but their admiration was quickly adopted by consumers who wanted to test it for themselves, pushing the smartphone app to the No.1 spot on the Apple App Store.
Despite it being a lesser-known brand, DeepSeek AI was found to show performance levels on par with OpenAI-o1, and it was also able to achieve this at a fraction of the cost, showcasing just how capable the tool was of using inferior hardware. This sets it up to expand and outperform other tools with further investments into DeepSeek AI capabilities. DeepSeek AI’s development potential for 2025 shows it is not done with its expansion.
The fact that the company is Chinese-run does bring in concerns of security similar to those surrounding ByteDance’s TikTok app and now, RedNote. For the most part, DeepSeek provides similar responses to questions the way Gemini or other mainstream AI do, however, you are also likely to run into roadblocks presented by the more restrictive content it is familiar with due to the region’s internet regulations.
When you realize what DeepSeek AI is and the parameters and content it has been trained on, you might see that the responses are equally conditioned.
Image: Pexels
Understanding the DeepSeek AI-Nvidia Connection
DeepSeek released its first AI large language model (LLM) in 2023, using Nvidia’s A100 graphics processor chips in the months before the U.S. restricted its sales in the region. The company then switched to Nvidia’s inferior H800 chips, which had not faced any sale restrictions in the region. If DeepSeek has also used Nvidia chips to build its AI, the same as OpenAI and other organizations, then why are Nvidia stocks crashing?
The reason is that DeepSeek was able to build its LLM within two months for less than $ 6 million, using the older, outdated technology. Companies like Google and OpenAI, for comparison, are expected to invest around $1 trillion in AI, a large amount of which is invested in purchasing the best, most advanced GPUs from Nvidia and other chipmakers like ASML.
As a result, investors began to fear that Nvidia’s chips and stocks had become overvalued. Many AI and energy stocks were affected on Monday when the DeepSeek AI gained attention. Nvidia took the biggest hit with a $600 billion drop in value. This 17% fall was the single-biggest one-day loss a company has faced in stock market history. Nvidia’s stock went back up by 9% just the next day, but the concerns surrounding its relevance have not been eliminated.
Companies like ASML and Broadcom took a hit due to their presence in the semiconductor industry, while GE Vernova and Vistra, two energy providers, were also affected.
How to Download DeepSeek AI
If you want to explore the DeepSeek AI’s capabilities for yourself, it is not hard to do. The DeepSeek R1 app is available globally on both the Apple App Store and Google Play Store, and you can create an account for yourself just like you would with any other app.
You can also access DeepSeek on your desktop through the company’s website. Setting up an account with your email address or phone number is easy but you can also directly log in via Google.
If you try to use the website right now, you might receive an unsettling notification “Due to large-scale malicious attacks on DeepSeek’s services, registration may be busy. Please wait and try again. Registered users can log in normally. Thank you for your understanding and support.”
It may be best to wait out the website’s return to normalcy before trying to create an account.
How to Use DeepSeek Locally
If you want to run DeepSeek -R1 locally on your system, you can download the Ollama tool and access it through its interface.
- Install Ollama from the website
- Open the terminal and type “ollama run deepseek-r1:1.5b” to instruct it to download the model.
- This command is for the smaller model, however, if your hardware supports it, you can run larger models by replacing the 1.5b with 7b, 8b, 14b, 32b, or 70b
- Check the installation using the “ollama list” command to ensure the model is available
- Launch the model using the “ollama run deepseek-r1” command and you should be able to use the model locally
If you have access to an Amazon Web Services (AWS) account and the Amazon Bedrock Custom Model, you can also use the DeepSeek-R1 Distilled Llama models through it.
There are many growing concerns around DeepSeek and its threat to US security as well as competition to US businesses. Despite that, the quality of the technology and the DeepSeek AI’s overall capabilities are impressive, to say the least.