A few introductory words about DeepSeek
Recently, the launch of the Chinese chatbot DeepSeek has caused quite a stir in the technology market.
The DeepSeek-R1 language model debuted on January 20, 2025, and quickly gained popularity, becoming one of the most downloaded applications in the Google Play and Apple App Store1.
This rapid growth surprised the industry, as competing AI models required many months to achieve a similar user base. In contrast, DeepSeek almost immediately became a competitor to giants like ChatGPT and Gemini.
However, this success was not without controversy.
The tool’s success triggered a significant “earthquake” on the American stock market, leading to declines in the stocks of these tech giants2. Nvidia was the biggest loser, with its shares dropping by as much as 17% on January 27, 2025, resulting in a loss of nearly $600 billion in market value in just one day3.
At the same time, concerns arose regarding censorship implemented in the model. Just as swiftly as the tool itself gained traction, reports circulated that the revolutionary chatbot avoids answering questions on topics sensitive from the perspective of the Chinese government.
Due to the above, some experts are already predicting the possibility of a DeepSeek ban in the United States, similar to what TikTok experienced4.
However, there is also a significant group of supporters involved in the DeepSeek debate. They particularly praise: the low-cost method of training the model while achieving results comparable to ChatGPT.
While Western companies like OpenAI and Google invest hundreds of millions of dollars in developing their chatbots, Chinese creators have achieved a similar effect for only $6 million.
As Forbes indicates5, the biggest winner in this whole story is not DeepSeek itself, but open-source artificial intelligence.
What is Open Source AI?
Open Source AI refers to projects, tools, and libraries related to artificial intelligence that are publicly available under open-source licenses. This means that their source code is accessible to anyone who can view, modify, and distribute it.
Similar to other IT systems, artificial intelligence can also be distributed under the principles of open-source.
This type of solution offers several benefits:
Open Source AI or Closed AI?
According to the definition provided by the Open Source Initiative, open-source artificial intelligence refers to AI systems made available under terms that guarantee complete freedom of use, without the need to obtain permission from their creators.
It offers the ability to analyze and modify the source code and allows sharing it with others, both in its original and modified forms.
Open-source models thus provide users with full control over the technology. They enable customization and the building of unique solutions based on them.
Closed AI technology operates on the opposite principles. Its source code and algorithms remain inaccessible, meaning that users cannot modify or develop them.
Proponents of this approach argue that keeping the code closed helps protect user privacy and prevents potential technological abuses.
However, both sides use the same argument.
In one article from Forbes, I found a quote from the co-founder and CEO of Labelbox, who notes that:
“(…) in today’s world, maintaining innovation in closed software is extremely difficult. Most key AI technologies are already open source, which influences the emergence of Open Source AI as a dominant trend in the industry.”
In a similar tone regarding DeepSeek, Marc Andreessen – co-creator of the Mosaic and Netscape Navigator browsers – stated:
“(…) it is one of the most amazing and impressive breakthroughs I have ever seen – and as open-source software, it is a great gift to the world.”
Open models also present an attractive alternative for companies seeking cost-effective solutions. The training cost of the DeepSeek-R1 model, which is only $6 million as mentioned earlier, seems negligible compared to the billion-dollar investments by OpenAI, Google, and Anthropic in developing closed models.
Considering the above, Open Source AI will undoubtedly continue to shake up the market…
Other Open Source AI Projects
The landscape of Open Source AI projects is continually evolving. The direction of progress in this technology may be influenced by several key players. In addition to the aforementioned DeepSeek, noteworthy projects include:
BLOOM
BLOOM (BigScience Large Open-science Open-access Multilingual Language Model) is a language model based on transformer architecture, created by the BigScience team.
It is one of the largest open-access AI models, comparable to GPT-3, but designed with an emphasis on openness, transparency, and scientific collaboration. The entire model and its parameters are publicly available. It serves as a model example of progress based on the cooperation of specialists.
Importantly, BLOOM is a multilingual model, which increases its utility across various regions and applications. It supports over 40 languages (including Polish) as well as many programming languages.
It can be used for text generation, translations, summarizations, data analysis, and many other natural language processing tasks.
The largest version of BLOOM has 176 billion parameters, making it one of the largest NLP (Natural Language Processing) models in the world.
Stable Diffusion
Stable Diffusion has revolutionized the space of text-to-image generation, providing an alternative to proprietary models. Developed by Stability AI and released as open-source, it allows anyone to download, modify, and use it without restrictions.
The model employs a diffusion technique, starting from random noise and gradually “denoising” it to produce an image that matches the provided description. It operates on neural networks and has been trained on vast datasets containing text-image pairs.
Users can create high-quality images based on textual descriptions, making it a valuable tool for artists, marketers, and content creators.
Unlike models such as DALL·E, it can be run for free on one’s own computer.
TensorFlow and PyTorch
TensorFlow and PyTorch are the two most popular frameworks for machine learning and deep learning. Both enable the building, training, and deployment of artificial intelligence models, utilizing neural networks for data processing.
TensorFlow was created by Google and is often used in large production projects due to its scalability and optimization for cloud and mobile environments.
PyTorch was developed by Meta (formerly Facebook) and is highly popular in academic settings and among researchers. It is known for its intuitive syntax, resembling standard Python code, which facilitates rapid experimentation and prototyping of new models.
OpenAI API*
Although OpenAI’s models, such as GPT-4 and DALL·E, are not fully open source, their open API allows for integration with a variety of tools.
This approach enables developers to build applications based on natural language processing, image generation, and data analysis. At the same time, it eliminates the need for having one’s own infrastructure to train large models. As a result, even small companies and independent developers can access technology that was previously available only to corporations and research institutions.
Additionally, OpenAI supports the development of the open-source community by publishing client libraries, implementation examples, and documentation that facilitates integration. Developers can also create their own interfaces, extensions, and solutions that utilize the OpenAI API and share them within the open ecosystem.
This hybrid approach combines the benefits of Closed AI technology with transparent collaboration among specialists. It allows for the development of innovative projects and broad applications of AI across various industries—from business process automation to education and creative content generation.
Open Source AI in CRM Systems
Open source artificial intelligence can also successfully find applications in CRM systems, especially those with open source code.
Open Source AI + Open Source CRM is an excellent way to achieve complete independence from technology vendors. Their open source code can be fully customized to fit the specifics of your business. Such solutions can be freely developed independently or with the support of the community or an experienced implementation company.
The advantage of using Open Source AI and Open Source CRM also lies in cost savings. Implementing AI in closed systems often involves high licensing fees and limited integration capabilities. In the case of open source, it may turn out that such costs are entirely avoided. Advanced features of Open Source technology (in most cases I know of) can be utilized without paying any subscription fees.
A few recommended applications of AI in CRM systems include:
The Future of Open Source AI
Looking ahead, the trend of democratizing AI through open source projects is likely to continue. Open Source AI is gaining significance as a trusted and robust solution for businesses.
Experts from IBM6 predict further development in this area, indicating that by 2025 we can expect:
Moreover, the success of tools like DeepSeek has contributed to a growing interest in Open Source AI technology from venture capital as well. The influx of significant financial resources strongly suggests that this trend will only intensify.
Personally, I believe that in the new year we can anticipate a wave of innovative open projects that will challenge the current industry leaders. The focus will likely shift towards ethical AI development. Following the controversy surrounding the censorship of the DeepSeek-R1 model, I expect the open source community to address concerns regarding the transparency of their proposed solutions. This approach may also result in a migration of IT professionals from large tech companies to startups developing new Open Source models.
All these factors should be crucial for the sustainable development of open artificial intelligence, which I wholeheartedly support.
- https://wydarzenia.interia.pl/zagranica/news-deep-seek-podbija-rynek-amerykanscy-giganci-traca-przewage-b,nId,7901727 ↩︎
- https://tvn24.pl/biznes/rynki/usa-chinski-deepseek-wywolal-ostra-przecene-firm-ai-na-wall-street-st8280759 ↩︎
- https://wydarzenia.interia.pl/zagranica/news-deep-seek-podbija-rynek-amerykanscy-giganci-traca-przewage-b,nId,7901727 ↩︎
- https://www.money.pl/gospodarka/tiktok-zablokowany-w-usa-170-milionow-uzytkownikow-bez-dostepu-surowe-kary-za-omijanie-zakazu-7115890308614976a.html ↩︎
- https://www.forbes.com/sites/kolawolesamueladebayo/2025/01/28/the-biggest-winner-in-the-deepseek-disruption-story-is-open-source-ai/ ↩︎
- https://www.ibm.com/think/news/2025-open-ai-trends ↩︎