DeepSeek has gained significant attention recently following the launch of its new R1 model, which is claimed to be on par with ChatGPT. The launch caused a stir in the market, leading to stock drops for major companies like Nvidia and OpenAI. DeepSeek’s iOS application has also become the most downloaded free app in the U.S., surpassing OpenAI’s app at the top.
Is DeepSeek-R1 actually better than ChatGPT? And more importantly—does it code better?
Nvidia has dominated the artificial intelligence (AI) chip market, controlling approximately 80% of it. The company sells the world’s most expensive AI chips—graphics processing units (GPUs). For Nvidia to maintain its success, it must remain the preferred choice for companies' AI needs, and customers must continue investing in its latest and most expensive chips.
However, at the end of January, something happened that raised questions about Nvidia’s ability to sustain its strong revenue growth in the long term. DeepSeek announced that it had trained its R1 model in just two months for under $6 million. In comparison, the largest U.S. tech companies have spent billions on Nvidia chips. As a result, Nvidia's stock plummeted by 17% in a single trading day, leading to a market value loss of nearly $600 billion.
DeepSeek is a Chinese AI company founded in 2023 that develops open-source large language models (LLMs). For those unfamiliar with LLMs, they are machine learning models trained on massive amounts of text data, enabling them to generate or predict human-like text.
DeepSeek-R1 is an AI model developed by DeepSeek and launched in January 2025. The model competes with—and in some cases surpasses—the reasoning capabilities of some of the world’s most advanced AI models but at a fraction of the operating cost, according to the company. DeepSeek-R1 is also open source under the MIT license, allowing for free commercial and academic use.
The fact that DeepSeek-R1 is open source means that the public has free access to both the model and its source code, which could have major implications for AI development going forward:
This is particularly important for Europe, as it provides an opportunity to join the “AI race.” Europe has long been dependent on AI models from U.S. companies such as OpenAI, Google DeepMind, and Meta. Open-source models allow European companies and researchers to build their own solutions without being locked into U.S. cloud services.
By being able to freely use, adapt, and host the model themselves, European entities can ensure data sovereignty, comply with strict privacy regulations like GDPR, and develop specialized solutions for multilingual and industry-specific needs. At the same time, DeepSeek’s API offers significantly cheaper access to advanced AI, enabling European startups and businesses to compete globally without the enormous costs typically associated with AI development.
One of the biggest differences between DeepSeek-R1 and ChatGPT lies in their focus. ChatGPT is primarily designed for conversation and focuses on narrow AI (task-specific intelligence). DeepSeek, on the other hand, aims to achieve artificial general intelligence (AGI).
Artificial general intelligence (AGI) is a hypothetical form of AI that can understand and learn any intellectual task in the same way a human can. The goal is to replicate the cognitive abilities of the human brain. The new model is designed to solve complex intellectual tasks at a level approaching human capacity, with a particular emphasis on enhanced reasoning skills. This makes it a significant contribution to the ongoing development of more powerful and flexible AI systems.
Previously, OpenAI set a standard for reasoning-based AI with its o1 model, which uses chain-of-thought techniques to break down problems into multiple steps. Through reinforcement learning (RL), the model can improve its strategies by adjusting based on reward systems, helping it identify mistakes and test alternative solutions when necessary.
DeepSeek-R1 builds on this approach by combining reinforcement learning with supervised fine-tuning, making it particularly well-suited for demanding logical and mathematical tasks. Test results show strong performance, with 79.8% on the AIME 2024 math test, 97.3% on the MATH-500 benchmark, and a Codeforces rating of 2,029, outperforming 96.3% of human programmers. In comparison, OpenAI’s o1-1217 scored 79.2% on AIME, 96.4% on MATH-500, and 96.6% on Codeforces.
In terms of general knowledge, DeepSeek-R1 achieved 90.8% accuracy on the MMLU benchmark, just one percentage point behind o1’s 91.8%. These results highlight DeepSeek-R1 as one of the most advanced open AI models on the market, demonstrating progress in reasoning and problem-solving for AGI development.
There is also a difference in the architecture of the models. DeepSeek-R1 utilizes a Mixture-of-Experts (MoE) approach, comparable to a team of specialized experts where only the most relevant ones are activated for each task. With a total of 671 billion parameters, DeepSeek-R1 activates only a subset (37 billion) for each request, increasing efficiency. This MoE architecture allows DeepSeek-R1 to optimize both performance and resource usage by dynamically adapting to different types of queries. In comparison, ChatGPT uses a traditional transformer model, where all parameters are activated for each task—ensuring consistent results but potentially lower efficiency.
Each model excels in different areas. DeepSeek-R1 particularly stands out in technical tasks, with impressive results in mathematics, making it valuable for tasks requiring precise technical solutions. ChatGPT, on the other hand, has a stronger ability to understand context and provide more nuanced responses across a broader range of topics.
For programmers, choosing the right AI tool is crucial. ChatGPT provides comprehensive coding assistance with detailed explanations, making it an excellent learning tool. DeepSeek-R1, however, focuses on faster code generation and more precise answers, making it particularly useful for quick and efficient solutions to specific programming challenges.
ChatGPT likely still provides better answers when you need to understand complex implementations. However, DeepSeek is more accurate in solving difficult tasks, as it focuses on delivering a single solution rather than an extensive answer with a lot of context and explanation. ChatGPT tends to provide multiple different suggestions for a problem, which can be both beneficial and disadvantageous, depending on the situation.
DeepSeek-R1 is still relatively new and has both advantages and drawbacks. During testing, we have experienced frequent server unavailability, which has been a drawback. However, with further refinement, DeepSeek could become a strong competitor to GPT.
DeepSeek remains free and open-source, giving developers the opportunity to create their own customized models. API access is also significantly cheaper than OpenAI’s competitor.
DeepSeek R1 is specifically designed for technical and structured tasks. It handles coding, mathematical reasoning, and logic-based questions with high efficiency, making it a great choice for many developers and researchers. If you seek precise and direct answers without unnecessary details, DeepSeek R1 delivers them efficiently.
ChatGPT o1, on the other hand, is more flexible. It performs well in creative writing, brainstorming, and open discussions, making it ideal for content production, research, and informal conversations. While it can handle technical topics, it tends to provide more detailed explanations, which can be useful for users who prefer more context.
