Global AI Model Call Volume Surpassed 129 Trillion Tokens

Chinese AI models have maintained a lead in weekly call volume over U.S. competitors for 21 consecutive weeks.

Updated on Sept. 21, 2026 in Artificial Intelligence

Bold flat-color editorial illustration featuring two stacks of circular data-wafers of differing heights, representing global artificial intelligence usage trends.
Global AI model usage reached 129 trillion tokens for the week ending September 20, 2026, as Chinese models continue to widen their lead over U.S. competitors. AI Illustration. Upload story photo >

Live Poll

Does the global growth of Chinese AI models make you more concerned about technology leadership?

Global AI large model call volume hit 129 trillion tokens for the week ending September 20, 2026. This reflects a total global increase of 1.57 percent week-over-week.

Why it matters

The data highlights a shifting landscape in artificial intelligence usage as Chinese models continue to dominate global call volumes compared to those developed in the United States.

DeepSeekV4.1-Flash secured the top spot in global call volume, processing 15.8 trillion tokens during the period. This model experienced a significant weekly volume increase of 219 percent.

The players

DeepSeekV4.1-Flash

This is a high-performance AI model that achieved the highest weekly call volume globally.

The details

Chinese models recorded 67.46 trillion tokens, marking a 10.28 percent increase week-over-week. Conversely, U.S. model volume dropped by 34.7 percent to 14.21 trillion tokens.

Timeline

  1. Global call volumes were measured during the week of September 14-20, 2026.

The Tech Race

This performance gap reflects an intensifying global arms race in large language model deployment and adoption. Chinese developers have sustained a lead for 21 weeks, indicating a structural shift in which regional ecosystems currently capture the highest volume of AI interactions.

Users may encounter varying response speeds or service availability depending on whether the AI tools they utilize are currently processing higher global loads. This data serves as a benchmark for developers gauging the scalability and infrastructure demands of current large-scale models.

The takeaway

The data underscores that global demand for AI is not uniform, with specific models like DeepSeekV4.1-Flash driving massive spikes in token volume. Analysts should watch these weekly metrics to identify which regional models are successfully scaling to meet enterprise and consumer needs.

Further reading

For more context on the current industry shifts, visit our Artificial Intelligence section.

Source note: This article includes information reported by Metal.

Live Poll

Does the global growth of Chinese AI models make you more concerned about technology leadership?