Four Chinese artificial ielligence models managed to take the first place in the global ranking of open source models; A topic that once again shows Beijing’s power in the competition of artificial ielligence with American gias.
According to LMARNA (AI -Model Evaluation Platform developed by researchers at Berkeley University), Kimi K2, Minimax M1, QWen 3 and Deepseek R1 models topped the list of the world’s best syhetic artificial ielligence models. These models succeeded in launching the products of companies such as Google (GEMA 3-72B) and Llama 4-Maverick.
The modeling of models in LMARNA is based on the comparison of the answers in the actual ieractions of the users and the top model is selected by mass voting. This is the same platform that companies like Openai and Google also use to measure their models.
In this ranking, the Kimi K2 model, the Chinese startup product, recely released on July 9th. This model has gained popularity due to “humorous and natural” responses and meal performance in human conversations.
Subsequely, the optimized version of the Deepseek R1 model, developed by Hangzhou’s startup, came in second. This model has been admired by the ability in “multi -stage and reasonable conversations”.
These successes come as Chinese models, especially in the open source, are competing with traditional Western actors at a remarkable speed. Analysts believe that as the trend coinues, China could play a more promine role in the global market of LLM models.
