📊 AI Model Usage Rankings

Which AI models are actually being used the most? Rankings based on real token consumption data from OpenRouter. Updated weekly. Data reflects actual developer and business usage patterns.

Top Models by Token Volume

Weekly token consumption across OpenRouter · Week of August 11, 2026

#ModelProviderTokens/WeekGrowth
1DeepSeek V4 FlashDeepSeek8.83T+570%
2Hy3Tencent8.05T+67%
3DeepSeek V4 Flash 0423DeepSeek5.88T+19%
4MiMo-V2.5Xiaomi5.39T+14%
5GPT-5.6 LunaOpenAI4.43T+128%
6GLM 5.2Z.AI3.49T+21%
7DeepSeek V4 ProDeepSeek2.61T+20%
8Nemotron 3 UltraNVIDIA2.41T+1%
9Gemini 3.6 FlashGoogle2.33T+446%
10Laguna S 2.1Poolside1.87T+101%
11Claude Opus 4.7Anthropic1.5T+85%
12GPT-5.5OpenAI1.2T+45%
13Qwen 3.7 MaxAlibaba1.1T+210%
14Gemini 3.1 ProGoogle0.9T+180%
15Kimi K2.6Moonshot AI0.8T+320%

Market Share by Provider

DeepSeek: 28%
OpenAI: 22%
Google: 18%
Anthropic: 12%
Z.AI: 6%
Tencent: 5%
Xiaomi: 4%
Others: 5%

Key Insights

  • DeepSeek dominates usage — 3 of top 10 models by volume, driven by aggressive pricing
  • Flash models win — cheap, fast models account for 60%+ of all tokens
  • Chinese models surge — Tencent Hy3, Xiaomi MiMo, Z.AI GLM growing rapidly
  • Google Gemini 3.6 Flash — fastest growing Western model (+446%)
  • Price matters most — the cheapest models get 10x more usage than the "best" ones