Chinese-built large language models now dominate the world’s largest LLM API aggregation platform, OpenRouter, accounting for 61% of total token consumption, according to data published on February 24, 2026.
The top three most-used models on the platform were all developed by Chinese AI labs. MiniMax M2.5 claimed the top spot with 2.45 trillion tokens consumed in a single week, a 197% increase from the prior week. Moonshot AI’s Kimi K2.5 followed with 1.21 trillion tokens, though its usage declined 20% week-over-week. Zhipu AI’s GLM-5 placed third with 780 billion tokens after surging 158%. DeepSeek V3.2 ranked fifth among all models on the platform. Together, Chinese models accounted for 5.3 trillion of the 8.7 trillion tokens consumed by the platform’s top 10 models.
The surge in MiniMax M2.5 usage was driven in part by free promotional access. Kilo Code, an AI-powered coding tool, offered the model at no cost for a week starting February 12. Cline, another developer tool, ran a similar promotion. MiniMax M2.5 scores 80.2% on SWE-Bench Verified, placing it within a point of Anthropic’s Claude Opus 4.6 at 80.8%.
Programming has become the single largest category of token consumption on OpenRouter, growing from 11% to over 50% of total tokens through 2025. Agent-driven workflows, in which models autonomously execute multi-step tasks, now account for more than half of all output tokens on the platform. OpenRouter COO Chris Clark stated that Chinese open-weight models are disproportionately heavy in agentic flows run by U.S. firms. All three top-ranked Chinese models this week are optimized for coding and agent automation.
Cost remains a defining factor. MiniMax M2.5 charges $0.30 per million input tokens and $1.10 per million output tokens. GLM-5 is priced at $0.30 per million input tokens and $2.55 per million output tokens. By comparison, Claude Opus 4.6 costs $5 per million input tokens and $25 per million output tokens, roughly 10 to 20 times more expensive.
OpenRouter’s total weekly token consumption has reached approximately 12.1 trillion, a 12.7-fold increase from a year ago. The platform, founded by former OpenSea CTO Alex Atallah, provides a unified API gateway to more than 400 models from over 60 providers. Andreessen Horowitz partner Martin Casado estimated that roughly 80% of startups using open-source AI stacks are running Chinese models. MIT Technology Review reported that adoption of Chinese open-weight models is accelerating in Silicon Valley.
