DeepSeek is a Chinese AI startup with a chatbot after it's namesake. Founded in 2023 by Liang Wenfeng, headquartered in Hangzhou, Zhejiang, DeepSeek is backed by the hedge fund High-Flyer.
DeepSeek's mission centers on advancing artificial general intelligence (AGI) through open-source research and development, aiming to democratize AI technology for both commercial and academic applications. The company focuses on developing open-source large language models (LLMs) that rival or surpass existing industry leaders in both performance and cost-efficiency.
Here's all the things you need to know about this new player in the global AI game.
What are DeepSeek's AI models?
DeepSeek-V3: Released in late 2024, this model boasts 671 billion parameters and was trained on a dataset of 14.8 trillion tokens over approximately 55 days, costing around $5.58 million. Benchmark tests indicate that DeepSeek-V3 outperforms models like Llama 3.1 and Qwen 2.5, while matching the capabilities of GPT-4o and Claude 3.5 Sonnet. Its architecture employs a mixture of experts with a Multi-head Latent Attention Transformer, containing 256 routed experts and one shared expert, activating 37 billion parameters per token.
DeepSeek-R1: Released in January 2025, this model focuses on logical inference, mathematical reasoning, and real-time problem-solving. It was trained using reinforcement learning without supervised fine-tuning, employing group relative policy optimization (GRPO) to enhance reasoning capabilities. This model achieves performance comparable to OpenAI's o1 across various tasks, including mathematics and coding.
How do I get access to DeepSeek?
DeepSeek's AI models are available through its official website, where users can access the DeepSeek-V3 model for free. Additionally, the DeepSeek app is available for download, providing an all-in-one AI tool for users.
Here's a deeper dive into how to join DeepSeek.
How does it compare to other models?
DeepSeek's AI models are distinguished by their cost-effectiveness and efficiency. For instance, the DeepSeek-V3 model was trained using approximately 2,000 Nvidia H800 chips over 55 days, costing around $5.58 million — substantially less than comparable models from other companies. This efficiency has prompted a re-evaluation of the massive investments in AI infrastructure by leading tech companies.
As a reference, let's take a look at how OpenAI's ChatGPT compares to DeepSeek. For example, OpenAI keeps the inner workings of ChatGPT hidden from the public. DeepSeek, in contrast, embraces open source, allowing anyone to peek under the hood and contribute to its development.
This fosters a community-driven approach but also raises concerns about potential misuse.
Both excel at tasks like coding and writing, with DeepSeek's R1 model rivaling ChatGPT's latest versions. However, DeepSeek's affordability is a game-changer.
Developed at a fraction of the cost, it demonstrates that cutting-edge AI doesn't have to break the bank.
ChatGPT offers a free tier, but you'll need to pay a monthly subscription for premium features. DeepSeek, on the other hand, believes in democratizing access to AI. This has fueled its rapid rise, even surpassing ChatGPT in popularity on app stores. Giving everyone access to powerful AI has potential to lead to safety concerns including national security issues and overall user safety.
ChatGPT is a complex, dense model, while DeepSeek uses a more efficient "Mixture-of-Experts" architecture. This allows it to punch above its weight, delivering impressive performance with less computational muscle.
DeepSeek operates under the Chinese government, resulting in censored responses on sensitive topics. ChatGPT, while moderated, allows for a wider range of discussions. This raises ethical questions about freedom of information and the potential for AI bias.
DeepSeek's arrival has sent shockwaves through the tech world, forcing Western giants to rethink their AI strategies. However, its data storage practices in China have sparked concerns about privacy and national security, echoing debates around other Chinese tech companies.
ChatGPT and DeepSeek represent two distinct paths in the AI environment; one prioritizes openness and accessibility, while the other focuses on performance and control. Their contrasting approaches highlight the complex trade-offs involved in developing and deploying AI on a global scale.
Is DeepSeek's technology open source?
Yes, DeepSeek has fully open-sourced its models under the MIT license, allowing for unrestricted commercial and academic use. This commitment to openness contrasts with the proprietary approaches of some competitors and has been instrumental in its rapid rise in popularity.
DeepSeek shows that open-source labs have become far more efficient at reverse-engineering. Any lead that U.S. AI labs achieve can now be erased in a matter of months. This is a serious challenge for companies whose business relies on selling models: developers face low switching costs, and DeepSeek’s optimizations offer significant savings.
What impact has DeepSeek had on the AI industry?
DeepSeek's advancements have caused significant disruptions in the AI industry, leading to substantial market reactions. The Chinese AI startup sent shockwaves through the tech world and caused a near-$600 billion plunge in Nvidia's market value.
The unveiling of DeepSeek’s V3 AI model, developed at a fraction of the cost of its U.S. counterparts, sparked fears that demand for Nvidia's high-end GPUs could dwindle.
This concern triggered a massive sell-off in Nvidia stock on Monday, resulting in the largest single-day loss in U.S. corporate history. The ripple effect also impacted other tech giants like Broadcom and Microsoft.
However, the panic proved short-lived. Nvidia's stock bounced back by almost 9% on Tuesday, signaling renewed confidence in the company's future. Experts point out that while DeepSeek's cost-effective model is impressive, it doesn't negate the crucial role Nvidia's hardware plays in AI development. In fact, the emergence of such efficient models could even expand the market and ultimately increase demand for Nvidia's advanced processors.
DeepSeek's breakthrough has seen mixed reactions. While Microsoft and OpenAI CEOs praised the innovation, others like Elon Musk expressed doubts about its long-term viability. Nvidia itself acknowledged DeepSeek's achievement, emphasizing that it aligns with U.S. export controls and shows new approaches to AI model development.
It’s clear that the crucial "inference" stage of AI deployment still heavily relies on its chips, reinforcing their continued importance in the AI ecosystem. The past few days have served as a stark reminder of the volatile nature of the AI industry. Disruptive innovations like DeepSeek can cause significant market fluctuations, but they also demonstrate the rapid pace of progress and fierce competition driving the sector forward.
How does DeepSeek recruit its talent?
DeepSeek focuses on hiring young AI researchers from top Chinese universities and individuals from diverse academic backgrounds beyond computer science. This strategy aims to diversify the knowledge and abilities within its models.
Has DeepSeek faced any challenges?
Yes, DeepSeek has encountered challenges, including a reported cyberattack that led the company to limit new user registrations temporarily. Despite these issues, existing users continued to have access to the service.
Additionally, tech giants Microsoft and OpenAI have launched an investigation into a potential data breach from the group associated with Chinese AI startup DeepSeek. The probe surrounds a look into the improperly acquired data from OpenAI's technology.
This investigation was sparked when Microsoft's security researchers observed what they believe to be DeepSeek-linked individuals extracting a substantial amount of data through OpenAI's application programming interface (API) in the fall of 2024.
The scale of data exfiltration raised red flags, prompting concerns about unauthorized access and potential misuse of OpenAI's proprietary AI models. Implications of this alleged data breach are far-reaching.
OpenAI, known for its ground-breaking AI models like GPT-4o, has been at the forefront of AI innovation. Its technology, accessible through APIs, has become a cornerstone for numerous applications across various industries. These APIs allow software developers to integrate OpenAI's sophisticated AI models into their own applications, provided they have the appropriate license in the form of a Pro subscription of $200 per month.
The potential data breach raises serious questions about the security and integrity of AI data sharing practices. As AI technologies become increasingly powerful and pervasive, the protection of proprietary algorithms and training data becomes paramount.
Are there concerns regarding DeepSeek's AI models?
Some sources have observed the official API version of DeepSeek's R1 model uses censorship mechanisms for topics considered politically sensitive by the Chinese government.
For example, the model refuses to answer questions about the 1989 Tiananmen Square protests and massacre, persecution of Uyghurs, or human rights in China. Additionally, there are fears that the AI system could be used for foreign influence operations, spreading disinformation, surveillance, and the development of cyberweapons for the Chinese government.
How has DeepSeek affected global AI development?
DeepSeek's rapid rise and technological achievements have prompted discussions about the global AI race, with some viewing its success as a "Sputnik moment" for the AI industry. This term suggests a significant shift in technological leadership, leading to reflections on the balance of innovation between China and Western countries
What are DeepSeek's future plans?
Currently, DeepSeek is focused solely on research and has no detailed plans for commercialization. This focus allows the company to concentrate on advancing foundational AI technologies without immediate commercial pressures.
Right now no one truly knows what DeepSeek’s long-term intentions are. DeepSeek appears to lack a business model that aligns with its ambitious goals. Unlike major US AI labs, which aim to develop top-tier services and monetize them, DeepSeek has positioned itself as a provider of free or nearly free tools — almost an altruistic giveaway. While this approach could change at any moment, essentially, DeepSeek has put a powerful AI model in the hands of anyone — a potential threat to national security and elsewhere.
DeepSeek’s rise demonstrates that keeping advanced AI out of the hands of potential adversaries is no longer feasible. As Anthropic co-founder Jack Clark noted, “DeepSeek means AI proliferation is guaranteed.”
More from Tom's Guide
