Introduction
Despite significant geopolitical tensions between the United States and China, Chinese open-source AI models are making inroads into the US market, outpacing numerous American programs and companies.
Open-Source AI Models vs. Proprietary Ones
Unlike the closed-source AI models, such as OpenAI’s ChatGPT and Google’s Gemini, which are fiercely guarded, Chinese open-source models allow programmers to customize software snippets to fit their specific needs. These models, offered by Chinese competitors like Alibaba and DeepSeek, are affordable—sometimes even free—and perform well.
According to a recent report by OpenRouter and US venture capital firm Andreessen Horowitz, the usage of open-source Chinese models has surged from a mere 1.2% at the end of 2024 to approximately 30% in August of this year.
The Impact of DeepSeek
DeepSeek’s high-performance, low-cost, open-source language model “R1” launched in January challenged the notion that top AI technology must come from US giants like OpenAI, Anthropic, or Google. It also served as a counterbalance to the US’s ongoing battle for AI dominance with China.
The US government acknowledges the potential of open-source technology. In July, the Trump administration’s “AI Action Plan” stated that the US needed “leading open-source models based on American values.” These models could become global standards, though US companies have taken the opposite route.
Meta, once a leader in the country’s open-source efforts with its Llama models, now focuses on closed-source AI. OpenAI, under pressure to recapture its nonprofit roots, recently released two “weight-open” models, slightly less flexible than open-source.
“Western Ideas”
Among Western giants, only Mistral from France remains committed to open-source, lagging far behind DeepSeek and Qwen in usage rankings. The American entrepreneur using Alibaba’s Qwen explained, “Western open-source offers aren’t as compelling.”
The Chinese government has promoted open-source AI technology despite doubts about its profitability. Mark Barton, OMNIUX’s technology director, said he was considering using Qwen but acknowledged that some clients might be uncomfortable interacting with a Chinese-made AI, even for specific tasks.
Given the current US administration’s stance on Chinese tech companies, risks persist. Barton stated, “We wouldn’t want to rely on a single model provider, especially one not aligned with Western ideas.” He added, “We don’t want to get caught in a bind if Alibaba were sanctioned or its use placed on a blacklist.”
A recent Stanford study suggested that “the very nature of open-model publications enables better scrutiny” of the technology. Gao Fei, BOK Health’s technology director in China, agrees: “Transparency and the shared nature of open-source are the best ways to build trust.”
Key Questions and Answers
- What are open-source AI models? Open-source AI models are technology platforms whose source code is publicly accessible, allowing developers to modify and distribute the code as they see fit.
- Why are Chinese open-source models gaining popularity in the US? These models are affordable, perform well, and offer customization options that appeal to American programmers.
- How has the US government responded to China’s open-source AI advancements? The US government acknowledges the potential of open-source technology and has called for leading open-source models based on American values. However, many US companies continue to develop closed-source AI.
- What are the risks associated with relying on Chinese open-source AI models? Given geopolitical tensions, there are concerns about potential sanctions or restrictions on using Chinese technology, which could leave companies vulnerable if they become overly dependent on a single provider.
