Skip to content

this month openrouter is chinese. two of the top 10 most-used models cost $0

pulse 	Illustration of person in purple hoodie swapping opus and gpt server racks for a red mimo-v2.5 unit in a data center.

openrouter top 5 by token volume – july 2026:

1. xiaomi mimo-v2.5 – 33.3t tokens
2. deepseek v4 flash – 25.4t
3. tencent hy3 (free) – 19.4t
4. minimax m3 – 14.3t
5. z-ai glm 5.2 – 14t

all five chinese. pricing: $0 to $0.60/m input. tencent hy3 ran free through july 21 – bulk of its 19.4t came from the promo window.

biggest mover: glm 5.2, up 241% after z-ai dropped mit-licensed open weights. 753b params, 40b active, $0.60/m input.

claude opus 4.8 at #8 with 7.39t. opus 4.7 at #9, down 22%. anthropic's two models combined still move less volume than mimo-v2.5 alone.

the default model for most builder workloads is no longer a frontier closed model. it's a chinese open-weight one priced 10–50x cheaper.

Ranking table of top 10 models by token usage on OpenRouter for July 2026, led by Xiaomi mimo-v2.5 at 33.3T tokens with 115% growth.

Stay in the loop

Get the latest AI news delivered to your inbox weekly

Thanks for subscribing!