DeepSeek V4.1-Flash-1017 Tops OpenRouter Weekly Chart with 33.6 Trillion Tokens
According to OpenRouter's latest weekly ranking (Oct 19-25), DeepSeek's V4.1-Flash-1017, updated on Oct 17, topped the global AI model leaderboard with 33.6 trillion tokens of weekly usage, up 801% week over week.
The version is based on V4.1 Flash released on Sep 10, retaining a 552B-parameter asymmetric MoE architecture with 8B activated input and 16B activated output parameters. It natively supports multimodality and focuses on code, automated agent, and vision tasks.
The second place was held by DeepSeek's older V4.1 Flash, with 18 trillion tokens this week, down 45% week over week, indicating developers are rapidly migrating to the new version. MiniMax M3.1 Flash ranked third with 13.7 trillion tokens. The top ten also included Zhipu AI GLM 5.3 Flash, Tencent Hy4 preview, and overseas models from OpenAI, Anthropic, and NVIDIA.
Total global AI model usage on the platform reached 188 trillion tokens last week, up 14.63% week over week. Chinese models accounted for 93.54 trillion tokens, up 18.14%, marking the 26th consecutive week that Chinese models surpassed U.S. models in weekly usage on the platform.
