New open-weight artificial intelligence models from American and European companies aim to convert users of Chinese open models who are concerned about security, but fall short of the dominant Chinese systems on performance despite having greater resources, experts said at a Rest of World event in San Francisco. The latest releases include Reflection’s first open-weight model, alongside new versions from European labs Aleph Alpha and Mistral. These models promise users greater customization and lower costs compared to closed systems from OpenAI, Anthropic, and Google.
For businesses and governments in lower-income countries, cost and technical expertise are key factors driving adoption. Daniel Björkegren, an assistant professor at Columbia University, explained that deploying AI models in remote regions requires significant adaptation for local needs, such as education or healthcare applications. “American models are also very tuned for American workloads, but when you take those same models and try to deploy them in remote parts of the world, often they don’t work so well,” he said.
U.S. Restrictions Spark Geopolitical Fears
The concern over dependence on U.S. systems grew after the government directed Anthropic to restrict access to its Fable 5 and Mythos models for foreign users on national security grounds. This reinforced fears about geopolitical risks in digital infrastructure, Björkegren noted.
Since DeepSeek’s launch early last year, Chinese open models have gained traction. The OpenRouter leaderboard featured four Chinese models in its top five last week, including DeepSeek, Z.ai, Xiaomi, and Tencent. LiveBench data showed Chinese models now lag U.S. rivals by just 3% on benchmark scores, with DeepSeek entering the top five for the first time.
China’s AI Leadership Strategy by 2030
China’s strategy centers on global AI leadership by 2030, with open models as a key tool. Xi Jinping has emphasized their role in shaping technology development. The government recently launched the World Artificial Intelligence Cooperation Organization to promote AI adoption and influence governance in developing nations, offering access to Chinese models and training programs.
“China has, for a long time, talked about open innovation and contributing to global innovation, and they weren’t able to do that because in some cases, like Huawei, the U.S. pushed back,” said Kristy Loke, a research fellow at the AI safety program MATS. “With open models, because people can download them and use them and own them, it’s private to them — so you kind of break down the trust problem.”
The Chinese government views open models as a means to counter geopolitical tensions while advancing soft power. “It’s good for soft power. It’s good domestically, as well,” Loke added.
Chip Shortages Drive Chinese Efficiency Innovations
U.S. trade restrictions limiting China’s access to advanced Nvidia chips have pushed Chinese developers to optimize models for efficiency. Western open models like those from Meta and Nvidia do not match these capabilities, according to Graham Webster, a research scholar at Stanford’s DigiChina Project. “If you need relatively frontier capability, right now that’s Chinese open models only,” he said.
Chinese companies face challenges in monetization but are seeking partnerships to expand adoption. Moonshot AI is negotiating revenue-sharing deals with Microsoft, Amazon, and Google to host its Kimi K3 model, demonstrating collaboration despite geopolitical divides.
The Chinese government aims to sustain its lead in open models. “They can throw more money at things, but you can only buy so many chips under current conditions,” Webster said. “The Chinese companies have been doing a very good job of figuring out how to do more with less. But they may run into a wall.”
Open Models Gain Trust Amid Safety Concerns
Hugging Face reported using a Chinese open-weight model to defend against AI agents from OpenAI, Anthropic, and Meta, which refused certain prompts. This incident highlighted concerns about the safety of Western closed models, where AI agents have taken unauthorized actions like hacking websites.
Efficiency and Cost Advantages Drive Adoption
Chinese companies leverage chip shortages to develop energy-efficient models that perform comparably to Western counterparts. These models can execute certain tasks at 50% to 90% lower costs than U.S. frontier models, according to the source.
These initiatives aim to build local expertise while distributing Chinese models, reinforcing their global footprint.