⬤ Kimi K2.5 has claimed the top spot on OpenRouter's LLM leaderboard based on actual usage numbers. The platform tracks real activity since it routes AI requests across multiple providers through one unified API, giving a clear picture of which models people actually use in practice.
⬤ The latest weekly data puts Kimi K2.5 at roughly 856 billion tokens processed, beating out Gemini 3 Flash Preview's 796 billion and Claude Sonnet 4.5's 711 billion. DeepSeek V3.2 sits in fourth with around 626 billion tokens. The charts show total token volume climbing steadily heading into this most recent measurement period.
⬤ These rankings show real demand rather than theoretical performance scores. Since OpenRouter handles routing between providers and includes automatic fallback systems for reliability, the token counts directly translate to how much each model is getting used on the platform.
⬤ The leaderboard reveals how usage is distributed across OpenRouter's available models, with the numbers indicating which AI systems developers and users are choosing most often.
Eseandre Mordi
Eseandre Mordi