Artificial intelligence is becoming more and more open source over time.
With the release of a potent large language model (LLM) in January 2025 that exhibits “remarkable reasoning capabilities” and performs on AI reasoning challenges on par with OpenAI o1, DeepSeek is a new player in the sector.
With its business model, DeepSeek, a Chinese startup established by Liang Wenfang in 2023, shows that open-source AI is not only competitive with proprietary AI solutions but may even outperform them in certain areas, particularly the contentious area of cost.
Can DeepSeek defeat OpenAI, though? This firm from Hangzhou will be a major force in the generative AI market for the foreseeable future if R1 is any indication. Here’s why.
All of the Current Knowledge Regarding DeepSeek R1
Benchmarks show that the open-source DeepSeek R1 AI model performs similarly to OpenAI in areas involving logic, coding, and mathematics. The model can be accessed using an API or the company’s official website.
In order to improve the model’s response accuracy, R1 has been trained utilizing methods like reinforcement learning and chain of thought reasoning, according to DeepSeek’s study report.
In addition to being open source, R1’s minimal API fee is one of its main selling features. For $0.14 per million input tokens and $2.19 per million output tokens, DeepSeek R1 is accessible through an API.
In comparison, R1 is incredibly less expensive than running O1 with comparable performance, costing $15 per million input tokens and $60 per 1 million output tokens.
Key Performance Metrics: R1 vs. O1
In what ways does R1 differ from O1? R1 performs similarly to O1 on a variety of coding and mathematical tasks, according to the performance measures described in DeepSeek’s study paper.
The closeness of R1 and O1’s overall performance against one another is what makes these scores noteworthy. Scores are about the same on AIME 2024, MATH-500, Codeforces, MMLU, and SWE-bench.
The GPQA Diamond, a series of multiple-choice questions created by subject-matter experts in chemistry, biology, and physics, is the only benchmark where we observe a slight deviation in performance.
Could OpenAI Be Threatened by R1?
Although DeepSeek R1 is excellent at offering a high-performance, open-source LLM at a cheap price, OpenAI will be hard to replace because the LLM industry is so crowded.
R1 is undoubtedly attracting a lot of interest, but it faces competition from goods with distinct points of uniqueness. For example, Claude’s safety-focused outputs, Grok’s less stringent content regulation, and OpenAI’s concentration on agentic AI all provide distinct distinctions that DeepSeek hasn’t yet properly defined.
However, OpenAI’s extensive commercial connections make it hard to compete with. For example, OpenAI launched The Stargate Project in January 2025, a $500 billion investment over the next four years from investors SoftBank, OpenAI, Oracle, and MGX.
R1 does, however, threaten OpenAI in one area: cost. Businesses may be enticed to choose R1 as a more affordable option if they find O1 to be unaffordable. After all, a lot of businesses find it difficult to cover the expenses of hiring and maintaining LLMs.
Professional Responses to DeepSeek R1 Thus Far
The release of DeepSeek R1 has been well received by the tech and AI communities, with many applauding the solution for its low cost and open-source methodology.
Nvidia’s DeepSeek project is aiming to maintain OpenAI’s original mission of open, frontier research. HouseTrip founder Arnaud Bertrand predicts DeepSeek will be a significant competitor to OpenAI.
China’s Deepseek has developed a model that matches or even surpasses OpenAI’s latest model, O1, on various benchmarks, at just 3% of the price. This is a significant blow to OpenAI, as it provides a new high-performance model for researchers to compete with, alongside other top performers like Llama 3.
The Bottom Line
DeepSeek has the ability to improve the AI ecosystem and reduce the barrier of entry for LLM. It remains to be seen if it can compete with OpenAI. The R1 release emphasizes how quickly generative AI is developing and how we cannot afford to ignore new companies that are trying to disrupt the market.
