OpenRouter

OpenRouter is an API service that gives developers access to “hundreds of AI models through a single API endpoint” and handles fallbacks between providers automatically.¹ OpenRouter, Inc. says it started in early 2023 “as the first LLM marketplace”; its About page lists more than 600 trillion tokens a month, more than 10 million users, more than 80 providers and more than 500 models Vendor-stated.² It passes through the underlying providers’ prices with “no markup on inference pricing” and charges a fee when customers buy credits instead; customers who bring their own provider keys pay a 5% fee on usage above a plan-dependent free allowance.³

Because requests to many models pass through one service, OpenRouter can count tokens, the units of text a model reads and writes, by model, and it publishes those counts as public rankings, licensed under CC BY 4.0. The rankings count prompt and completion tokens together, aggregated in daily UTC buckets per model variant; they cover only traffic routed through OpenRouter, and each model’s tokens are counted in its own tokenizer.⁴⁵ In the seven days to Oct 12, 2026, DeepSeek V4.1 Flash led with 40.53 trillion tokens, and the top five models were all from Chinese labs Measured, as ranked in Most-Used AI Models on OpenRouter by Weekly Tokens.⁴

With the venture firm a16z, OpenRouter also published a study of more than 100 trillion tokens of its traffic in December 2025. Over the year to November 2025, it found, Chinese open-source models averaged about 13.0% of weekly token volume across all models, from a weekly low of 1.2% in late 2024 to nearly 30% in some weeks Measured.⁶ Chinese Open-Weight Models: Downloads, Derivatives and Token Share sets those shares beside download counts from Hugging Face, and notes the study’s own caveat that one platform leaves out enterprise, local and internal deployments.⁶