Chinese large language models (LLMs) have held the top six spots in global usage rankings on OpenRouter, a leading AI model marketplace, for five straight weeks. This surge highlights a pivotal shift in the U.S.-China AI rivalry, fueled by aggressive free-tier pricing from firms like Alibaba and Xiaomi.
Usage Stats Surge
Last week, from March 30 to April 5, global AI model usage hit 27 trillion tokens, up 18.9% week-over-week. Chinese models processed 12.96 trillion tokens, a 31.48% jump, while U.S. models managed just 3.03 trillion, growing only 0.76%. Alibaba’s free Qwen3.6 Plus led with 4.6 trillion tokens, including over 1 trillion in a single day post-launch on April 2; its Qwen3.6 Plus Preview took third at 1.64 trillion.
Brains vs. Bodies Divide Blurs
Analysts describe the competition as U.S. strength in AI “brains” (chatbots, chips, LLMs) versus China’s edge in “bodies” (humanoid robots), but boundaries are fading. NVIDIA powers U.S. advantages in advanced chips, while labs like OpenAI, Anthropic, and Google lead benchmarks, yet China now builds competitive LLMs at lower costs. A BBC report notes the gap is “far less one-sided,” with Chinese open-source strategies accelerating iteration.
Broader Platform Trends
Beyond OpenRouter, Chinese models fill half or more of Hugging Face’s top trending spots, per platform head Jeff Boudier. Bayuegua Institute data shows China hosting 51 of the top 100 global AI firms, versus 37 in the U.S. Models like Qwen3, GLM-4.5, and Alibaba’s offerings dominate open-source charts.
Strategic Implications
Free access undercuts U.S. paid models, boosting adoption, but experts like Eurasia Group’s Ali Wyne predict niche wins over total dominance. China’s open-source ethos and volume edge contrast U.S. IP protections, potentially reshaping global AI dynamics. Whether usage yields commercial leads remains key to watch.
0 Comments