Western open-weight models still trail China's on performance, experts say
Experts speaking at a Rest of World event in San Francisco said a wave of new open-weight models from American and European companies aims to win over users of Chinese open models worried about security, but still falls short of the dominant Chinese systems on performance. The article notes that US startup Reflection released its first open-weight model last week, and that European labs Aleph Alpha and Mistral also released new open-weight models. It adds that four Chinese models were among the top five on OpenRouter's leaderboard last week, that LiveBench recently placed DeepSeek in its top five for the first time, and that top Chinese models trailed US rivals by just 3% on benchmark scores; Chinese open models can do some tasks 50% to 90% cheaper than leading American frontier models, though Chinese companies still face compute limits and monetization questions.
The article says open-weight models can be downloaded, used and owned locally, easing some governments' concern about depending on US frontier intelligence, while the lower cost of Chinese open models is described as a route to overseas markets and soft power.
A slew of 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.
Last week, American startup Reflection launched its first open-weight model, and European labs Aleph Alpha and Mistral also released new open-weight models. Open models such as these offer users greater customization and control than the closed models of OpenAI, Anthropic, and Google at significantly lower costs.
For businesses and governments in lower-income countries, in particular, cost is the primary factor, as well as the expertise needed to customize open models for their specific needs, including in areas such as education, healthcare, and farming, Daniel Björkegren, an assistant professor at Columbia University, said at the Rest of World event last week.
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.”Daniel Björkegren, assistant professor at Columbia University
#Reflection#Aleph Alpha#Mistral#DeepSeek#Z.ai#Tencent#Xiaomi#Moonshot AI#Kimi K3#Nvidia#Thinking Machines