A new Chinese artificial intelligence model, DeepSeek, has surged to the top of the Apple Store download charts, surprising investors and causing a dip in several tech stocks.
Launched on January 20, its latest version quickly drew praise from AI experts and captured the attention of the tech industry and beyond.
Former President Donald Trump remarked that this development serves as a “wake-up call” for American businesses, emphasizing the need to “compete to win.”
DeepSeek stands out for its reported ability to deliver results at a fraction of the cost of leading models like those from OpenAI, primarily due to its use of fewer advanced chips.
This revelation prompted Nvidia, a major player in chip manufacturing, to experience a staggering loss of nearly $600 billion in market value on Monday—marking the largest single-day drop in U.S. history.
Moreover, DeepSeek raises significant concerns regarding the U.S. government’s strategies to limit China’s technological ambitions, particularly since one of the key measures has involved restricting the export of advanced chips to the country.
In response, China has intensified its focus on AI, with President Xi Jinping designating it as a national priority. Start-ups like DeepSeek are vital as China transitions from traditional industries such as textiles and furniture to more advanced technology sectors like chips, electric vehicles, and AI.
So, what do we actually know about DeepSeek?
What is DeepSeek?
DeepSeek is a free AI-powered chatbot designed to offer an experience similar to ChatGPT. It tackles a variety of tasks, although opinions vary on how it stacks up against other options available.
It’s said to be on par with OpenAI’s o1 model, released last year, especially in areas like mathematics and coding. Like o1, DeepSeek operates as a “reasoning” model, generating responses step by step, mimicking the way humans approach problem-solving. This approach allows it to use less memory than its competitors, ultimately keeping operational costs down.
In keeping with the approach of other Chinese AI models like Baidu’s Ernie and ByteDance’s Doubao, DeepSeek is designed to avoid engaging with politically sensitive subjects. For example, when the BBC asked about the events at Tiananmen Square on June 4, 1989, DeepSeek declined to comment on the massacre, which is a highly sensitive topic in China.
The Chinese government’s censorship presents a significant obstacle for DeepSeek’s international AI ambitions. However, its foundational model seems to have been constructed from reliable sources, while an additional safeguarding layer ensures certain information is withheld.
DeepSeek reports that it has developed its technology at an impressively low cost, stating that their training budget was just $6 million (£4.8 million). This figure stands in stark contrast to the “over $100 million” noted by OpenAI’s CEO, Sam Altman, in relation to GPT-4.
The DeepSeek founder is said to have built up a significant collection of Nvidia chips, even though their export to China has been restricted since September 2022. Some experts estimate this stash, potentially around 50,000 units, enabled the development of a robust AI model by combining these high-end chips with less expensive alternatives.
On the same day DeepSeek’s AI assistant was named the most downloaded free app in the US on Apple’s App Store, the company encountered significant malicious attacks, leading to a temporary suspension of new registrations. Additionally, there were outages on the website on Monday.
Who is behind DeepSeek?
DeepSeek was established in December 2023 by Liang Wenfeng, who quickly made a mark with the launch of its first AI large language model the following year.
Details about Liang are relatively sparse. An alumnus of Zhejiang University, he holds degrees in electronic information engineering and computer science, but he has recently come into the public eye.
Liang was spotted at a meeting chaired by China’s premier, Li Qiang, signaling DeepSeek’s rising significance in the AI landscape.
In contrast to many American AI entrepreneurs who hail from Silicon Valley, Mr. Liang also brings experience from the finance sector. He serves as the CEO of High-Flyer, a hedge fund leveraging AI to analyze financial data for investment strategies, commonly referred to as quantitative trading. Notably, in 2019, High-Flyer became the first quantitative hedge fund in China to surpass 100 billion yuan (approximately $13 million) in funds raised.
In a speech in 2019, Liang shared his perspective on the issue, stating, “If the US can advance its quantitative trading sector, why shouldn’t China be able to do the same?”
He elaborated, “It’s often suggested there’s a one or two-year gap between Chinese and American AI, but the true divide lies between originality and imitation. If this dynamic doesn’t shift, China risks perpetually being in a follower’s role.”
When asked about the surprise generated by DeepSeek’s model among Silicon Valley insiders, he remarked, “Their astonishment reflects the emergence of a Chinese company as an innovator in the field, rather than merely a follower—a distinction that many Chinese firms are typically seen as.”
In the meantime, Australia’s science minister has expressed some concerns regarding the security of the app.
“There are a lot of questions that will need to be answered over time concerning quality, consumer preferences, and data and privacy management,” Ed Husic stated in an interview with ABC.
“I would advise caution. These types of issues require careful consideration.”
DeepSeek AI vs ChatGPT: A Price and Features Comparison
Here is the table format of the data for easy understanding:
| Benchmark (Metric) | DeepSeek V3 | DeepSeek V2.5 | Qwen2.5 (72B-Inst) | Llama3.1 (405B-Inst) | Claude-3.5 (Sonnet-1022) | GPT-4o (0513) |
|---|---|---|---|---|---|---|
| Architecture | MoE | MoE | Dense | Dense | – | – |
| # Activated Params | 37B | 21B | 72B | 405B | – | – |
| # Total Params | 671B | 236B | 72B | 405B | – | – |
| English | ||||||
| MMLU (EM) | 88.5 | 80.6 | 85.3 | 88.6 | 88.3 | 87.2 |
| MMLU-Redux (EM) | 89.1 | 80.3 | 86.0 | 88.2 | 87.8 | 86.5 |
| MMLU-Pro (EM) | 75.9 | 66.2 | 71.6 | 73.3 | 78.0 | 72.6 |
| DROP (3-shot F1) | 91.6 | 80.6 | 84.1 | 86.0 | 86.5 | 83.7 |
| iF-Eval (Prompt Strict) | 86.1 | 80.6 | 84.1 | 86.0 | 86.5 | 84.3 |
| GPAI-Diamond (Pass@1) | 59.1 | 41.3 | 49.0 | 51.1 | 65.0 | 49.9 |
| SimpleQA (Correct) | 24.9 | 10.2 | 19.1 | 21.0 | 28.4 | 18.2 |
| FRAMES (Acc.) | 73.3 | 65.4 | 69.8 | 70.0 | 72.5 | 80.5 |
| LongBench v2 (Acc.) | 48.7 | 35.4 | 39.4 | 36.1 | 41.0 | 48.1 |
| Code | ||||||
| HumanEval-Mul (Pass@1) | 82.6 | 77.4 | 77.3 | 77.2 | 81.7 | 80.5 |
| LiveCodeBench (Pass@1-COT) | 40.5 | 29.2 | 31.1 | 28.4 | 36.3 | 33.4 |
| LiveCodeBench (Pass@1) | 37.6 | 28.4 | 28.7 | 30.1 | 32.8 | 34.2 |
| Codeforces (Percentile) | 51.6 | 35.6 | 24.8 | 23.8 | 32.2 | 33.2 |
| SWE Verified (Resolved) | 42.0 | 22.6 | 23.8 | 24.5 | 50.8 | 38.8 |
| Aider-Edit (Acc.) | 79.7 | 71.6 | 65.4 | 63.9 | 84.2 | 72.9 |
| Aider-Polyglot (Acc.) | 49.6 | 18.2 | 7.6 | 5.8 | 45.3 | 16.0 |
| Math | ||||||
| AIME 2024 (Pass@1) | 39.2 | 16.7 | 23.3 | 23.3 | 16.0 | 9.3 |
| MATH-500 (EM) | 90.2 | 74.7 | 80.0 | 73.8 | 78.3 | 74.6 |
| CNMO 2024 (Pass@1) | 10.8 | 1.5 | 1.9 | 6.8 | 4.3 | 3.7 |
| Chinese | ||||||
| CLUEWSC (EM) | 90.9 | 90.4 | 91.4 | 84.7 | 85.4 | 87.9 |
| C-Eval (EM) | 86.5 | 79.5 | 86.1 | 61.5 | 76.7 | 76.0 |
| C-SimpleQA (Correct) | 64.1 | 54.1 | 48.4 | 50.4 | 51.3 | 59.3 |
When looking at DeepSeek and ChatGPT, both platforms excel in generating responses and tackling complex inquiries, but they have distinct differences. ChatGPT, created by OpenAI, boasts a diverse array of features, including the ability to create AI-generated images, engage with tools like Canvas, and offer multimodal interfaces for tasks such as image analysis. ChatGPT also allows for voice interaction, providing a more flexible experience, whereas DeepSeek focuses solely on text-based queries.
In terms of performance, DeepSeek-V3 and R1 hold their own against ChatGPT, particularly in handling technical questions and code generation. Both platforms feature web search capabilities, ensuring users access the most current and reliable information. Nevertheless, ChatGPT enhances its offering with customized GPTs, which enable users to tailor their AI experience for specific needs, greatly benefiting those looking for personalized AI assistance in their daily routines.
What sets DeepSeek apart is that it is entirely free to use, with no limits on the number of queries. While ChatGPT does offer a free version, advanced features require a subscription. Additionally, DeepSeek’s API costs are remarkably lower, making it an appealing choice for developers and businesses. For example, DeepSeek charges just $0.55 per million input tokens and $2.19 per million output tokens, which is substantially more budget-friendly compared to OpenAI’s API pricing of $15 and $60 for similar services.
DeepSeek’s Open-Source Advantage
One of the standout characteristics of DeepSeek is its open-source framework. Unlike ChatGPT, which operates on a proprietary system, DeepSeek’s R1 code is available to the public under the MIT License. This allows developers the freedom to use, modify, and distribute the code, fostering greater integration possibilities across various applications without facing licensing fees. The open-source model has attracted considerable interest, paving the way for enhanced innovation and customization.
DeepSeek is shaking things up in the AI landscape, even making waves in the U.S. tech scene. By early 2025, DeepSeek-V3 had claimed the title of the most-downloaded app on the iPhone App Store in the United States, marking an impressive milestone for a Chinese-developed application. This success has sparked discussions about how some U.S. tech firms, which have poured billions into AI advancements, are managing their resources.
What really sets DeepSeek apart is its efficiency. Despite operating on a budget that’s just a fraction of what giants like OpenAI or Google allocate for AI development, DeepSeek’s models are strikingly effective. Reports indicate that developing the DeepSeek-V3 model cost under $6 million—remarkably less than the billions spent by competitors. This success raises questions about the widely held belief that only substantial financial investments can lead to top-tier AI.
Who Comes Out on Top: DeepSeek or ChatGPT?
While both DeepSeek and ChatGPT provide powerful AI capabilities, they serve different user preferences. ChatGPT excels with its extensive feature set, including voice interaction, image creation, and custom GPTs, making it an excellent choice for users seeking a versatile, multimodal AI experience. In contrast, DeepSeek’s noteworthy performance in complex reasoning tasks, affordability, and open-source nature position it as the go-to option for users and developers in search of a straightforward, budget-friendly AI tool.
Navigate to DeepSeek: https://deepseek.com
Related Topics :
https://visionarydaily.in/chatgpt-unveiling-the-wonders-of-chatgpt/
