China’s AI Models Lead Global Token Calls for Eighth Week
China’s artificial intelligence large language models have maintained the global lead in token call volume for the eighth consecutive week, with Chinese models processing 18.81 trillion tokens in the week of June 15-21, 2026 — more than triple the 5.76 trillion tokens recorded by US models over the same period, according to data from OpenRouter, the world’s largest AI model API aggregation platform.
Global AI large model total call volume reached 46.7 trillion tokens for the week, marking nine consecutive weeks of growth, as reported by People’s Daily. China’s weekly token consumption grew 2.12% week-over-week, outpacing the US growth rate of 0.70%, further widening the gap between the two nations’ AI deployment scales.
Understanding Token Call Volume
A token is the fundamental unit of data processed by an AI large language model — representing words, syllables, or punctuation. Token call volume measures the total number of these units a model processes, directly reflecting real-world AI application scale and market adoption. Higher token usage indicates broader deployment and commercial viability.
For years, global AI competition centered on a “parameter race,” with companies vying to build larger models. However, the industry has shifted: parameter scale represents theoretical potential, while token usage is the key metric for commercial viability. China’s sustained lead signals that its AI industry has moved from research and development to large-scale deployment.
Market Leaders and Competitive Dynamics
DeepSeek-V4-Flash topped the domestic rankings with 4.94 trillion weekly tokens, up 12% week-over-week, holding the number one spot for five consecutive weeks. Xiaomi’s MiMo-V2.5 surged to second place with 3.94 trillion tokens (+10% WoW), while MiniMax M3 ranked third with 3.77 trillion tokens (-13% WoW). Tencent’s Hy3preview fell to fourth with 3.63 trillion tokens (-12% WoW), as reported by People’s Daily.
Industry analysts cited by Securities Daily noted that the data “not only marks China’s established global lead in AI application scale, but also reflects that the domestic AI large model industry has fully entered the deep-water zone of large-scale application with ‘value realization’ at its core.”
The Broader Token Economy
China’s AI momentum extends well beyond weekly rankings. According to the National Data Administration, China’s average daily token call volume exceeded 140 trillion tokens in March 2026, up from 100 billion in early 2024 — a 1,400-fold increase in just over two years, as reported by the Global Times.
ByteDance’s Doubao model now processes over 120 trillion tokens daily, doubling in three months and increasing thousandfold compared with two years ago, according to Tan Dai, president of ByteDance-backed Volcano Engine, as reported by China Daily.
Zhang Yi, CEO of iiMedia Consulting, told Securities Daily that China’s AI large models can lead globally long-term due to two core advantages: “rich real-economy scenarios and massive national-level traffic entrances” combined with “the open-source ecosystem with extreme cost control capabilities.”
Telecom Giants Enter the Token Market
China’s three major telecommunications operators — China Mobile, China Telecom, and China Unicom — have all launched “Token Factory” services, commoditizing AI computing power as a utility. As FreshFromChina reported, China Mobile launched a token computing service for individual users starting at ¥5.99 ($0.83 USD), while China Telecom introduced trial commercial packages for SMEs at ¥39.9 per month ($5.51 USD).
Pan Helin, a member of the Ministry of Industry and Information Technology’s advisory committee, explained to China Daily that “a ‘computing supermarket’ allows enterprises to purchase processing capacity on demand, while a ‘computing bank’ enables credit-based usage or even financing backed by computing assets. These new models essentially financialize computing power.”
Industry Maturation and Consolidation
The rapid growth in token consumption is accompanied by significant market consolidation. Data from Qichacha, cited by People’s Daily, shows that China now has approximately 13,000 AI large model enterprises, with 44.52% having registered capital of 50 million yuan or more. However, new registrations have slowed sharply — just 24 new companies in 2026 year-to-date, compared to 395 in all of 2025 — indicating a maturing market where capital-intensive barriers favor established players.
Wang Peng, an associate research fellow at the Beijing Academy of Social Sciences, told the Global Times that “in China, large models are quickly embedded into high-frequency platforms such as e-commerce, social media and public services, while aggressive pricing strategies, including free-tier access, have significantly lowered adoption barriers and expanded the user base.” He also noted that token consumption is structurally higher due to the rapid expansion of AI agents and multi-step workflows.
Divergent US-China AI Paths
The data highlights fundamentally different approaches to AI development. The US path, led by OpenAI, prioritizes artificial general intelligence breakthroughs with a closed-source API model and high pricing targeting the high-end market. China’s path emphasizes diversified models, open-source ecosystems, aggressive pricing, and deep integration with real-economy scenarios.
As ChinaBiz Insider noted, China’s global lead in token usage “proves that scenario-driven commercialization is a viable and successful development path,” with the industry shifting from technical parameter competition to large-scale commercial application.
What to Watch
Looking ahead, several factors will shape the trajectory of China’s AI token economy. The Ministry of Industry and Information Technology has outlined plans to pilot “computing banks” and “computing supermarkets” nationwide, with a goal by end of 2028 to establish a broadly accessible computing service system. Meanwhile, investment bank CICC projects that AI agent penetration could drive a fivefold increase in daily token consumption, which would further entrench China’s lead given its massive user base.
Key questions remain: how much of China’s token volume is driven by free and low-cost pricing versus genuine demand, and whether US models can regain ground through AGI breakthroughs or different commercialization strategies. The impact of export controls on advanced chips also continues to be a factor in China’s ability to maintain compute infrastructure growth.
What is clear is that China’s AI industry has crossed a critical threshold — from a technology follower to a market leader in global AI deployment scale, with implications that will resonate across the entire technology landscape.