The Rise of Chinese AI Models: A New Contender in the U.S.

Chinese artificial intelligence (AI) products are rapidly gaining popularity in the United States and around the world. The San Francisco-based chief technology officer at Mozilla, Raffi Krikorian, recently shifted to using the Chinese AI startup Moonshot’s Kimi K3 for daily tasks. He found it more responsive than the more expensive Claude Fable chatbot from Anthropic. The Kimi K3 model is a testament to the increasing appeal of Chinese AI systems, which are often more cost-effective and efficient.

As American companies like Coinbase look to cut costs by adopting Chinese AI models, this trend is causing frustration among some U.S. tech giants. Despite American-led restrictions to prevent China from accessing advanced technologies, the appeal of affordable and efficient Chinese AI models is undeniable. U.S. President Donald Trump’s administration accused Moonshot of controversial practices in building K3, sparking tensions between the two countries. However, Beijing dismissed these accusations.

Growing Capabilities of Chinese AI Models

In recent years, Chinese AI models have been closing the gap with U.S. counterparts. Startups like Z.ai and Moonshot launched advanced models that are nearly as capable as those from U.S. giants like OpenAI and Anthropic. For example, DeepSeek’s V4 model challenged the performance of OpenAI’s GPT and Google’s Gemini.

The advancement of these models is yielding practical results. Curt Meinhold, a technology executive from North Carolina, prefers using DeepSeek for business-related tasks. He notes that many people do not require the highest-end models from Anthropic when affordable alternatives are available. U.S. investment bank Goldman Sachs also highlighted the potential of Chinese AI models to meet the growing global demand for AI solutions.

Price and Accessibility

AI models, priced based on the number of tokens they process, present a significant cost advantage for consumers choosing Chinese options. Analysts like Alex Colville emphasize the cost-effectiveness of Chinese models, which offer similar capabilities at a reduced price compared to American models.

Challenges and Opportunities

Despite their growing popularity, Chinese AI models still fall short in some areas compared to their U.S. counterparts. Anastasios Angelopoulos, CEO of Arena, believes that while there is potential, these models need to improve to match the full spectrum of capabilities offered by American leaders.

In the U.S., there is eagerness among AI firms to reduce prices by exploring competitive, cost-effective models. As many Chinese AI systems are open-source, they offer opportunities for global expansion and adaptation. Lian Jye Su from Omdia notes the increased adoption of Chinese models as they continue to match the capabilities of closed-source American models.

Other U.S. policies may inadvertently favor Chinese models. For example, the export controls on Anthropic models last June allowed Chinese competitors to gain an advantage.

Global Reach and Strategic Implications

With strong state support, Chinese tech companies are integrating AI into various devices, including smartphones and AI glasses. President Xi Jinping emphasized the significance of open-source AI models at a technology summit in Shanghai, promoting their use globally.

Meanwhile, Chinese AI companies are raising funds to expand internationally, capitalizing on domestic competition and opportunities abroad. This fierce competition drives Chinese labs to outdo each other, adding pressure to the market.

Both the U.S. and China are pursuing widespread AI ecosystem adoption while safeguarding strategic technologies. However, economic sustainability remains a concern as demonstrated by Z.ai’s recent financial results. Despite these challenges, Chinese AI models are poised for further growth, including in the U.S. marketplace.

Krikorian recommends evaluating these models for any serious AI tasks, reflecting the shifting landscape in the AI industry.

Reported by AP’s Kelvin Chan and others. Patrick O’Brien contributed from Rhode Island.