Why Is DeepSeek Giving Away the best AI Model for Free?
I recently came across a new AI model from DeepSeek called R1, which directly competes with OpenAI’s O1 model. If you’re unfamiliar, OpenAI’s O1 is the world’s first reasoning model, meaning it takes more time to think before responding. This makes it ideal for research institutions, educational platforms, and tech companies that need advanced problem-solving capabilities.
One thing that really caught my attention was Deep Seek’s MIT Open License. For beginners like me, this license means: Free to use, Freedom to modify and distribute, and attribution required. Simply put, I can build new software using DeepSeek R1, sell it to other companies, and I don’t need to pay DeepSeek anything.
What’s even more exciting? The performance. Reports suggest that the DeepSeek V3 mixture-of-experts model delivers performance similar to OpenAI’s O1. However, I’m a bit skeptical about its real-world effectiveness. Some reports indicate that parts of DeepSeek’s model were trained using responses from ChatGPT, which raises questions about originality and performance reliability.
Overall, DeepSeek R1 looks promising, and I’m eager to see how it competes with OpenAI’s offering in the long run.
So, what’s the strategy at play? Why give DeepSeek model for free?
A free model could lead to mass adoption across the industry, including tech companies and developers. This widespread adoption allows the community to detect bugs early and contribute by developing new techniques to enhance the model.
This wide spread adoption leads all these companies and developers to stop using other paid APIs and models. Over time, the reduced funding could lead to a slowdown in innovation, with fewer resources available for creating next-generation models. The competitive landscape could become less intense, as smaller players relying on free models dominate the ecosystem, while the larger, closed-source companies bleed billions of dollars. This long-term financial strain could hinder advancements in AI and potentially stall progress in achieving transformative breakthroughs.
With DeepSeek’s resources combined with community-level adoption, the model could become one of the best in the world. Could it surpass OpenAI and Meta? Perhaps, or perhaps not.
How do they make money?
For a startup looking to integrate an AI model for customer support, investing in expensive graphics cards and deploying them on in-house servers isn’t always feasible due to high upfront costs. Instead, the smarter choice is to integrate AI APIs that are cost-effective and charge based on actual usage.
As the startup grows 10x, these AI APIs need to scale accordingly, simplifying infrastructure management and reducing operational complexity. APIs act like a magic pill — you plug them in, and they take care of scaling, updates, and even the latest AI advancements without any manual intervention.
DeepSeek offers such APIs, and the cherry on top? They are 90–95% more affordable than OpenAI’s.
But if it’s that simple, everyone should be using DeepSeek?
Challenges: Trust and Privacy
When it comes to AI adoption, Trust and Privacy are paramount, especially for businesses handling sensitive customer data. According to their private policy user data may be stored on secure servers located in China, which raises questions about compliance with regulations such as GDPR.
Moreover, the policy grants DeepSeek an unconditional, irrevocable license to reproduce, use, and modify user inputs and outputs, which could be a dealbreaker for organizations with stringent data protection policies. This lack of clarity, coupled with the potential for data sharing within their corporate group, adds to the skepticism surrounding DeepSeek’s suitability for enterprise-level applications. (This is more relevant for the API usage)
Links:
https://chat.deepseek.com/downloads/DeepSeek%20Privacy%20Policy.html
