Junghwan Lim, CEO of Korean artificial intelligence startup Motif Technologies, said the company used a $25 million Nvidia B200 GPU, which is backed by the government to develop its latest big language model, Motif 3. Motif 3 scored 47 points in the Artificial Analysis Intelligence Index, which is far above the average of similar models and is on par with Alibaba's Qwen 3.7 Max. Motif is exploring a new token-as-a-service business model that allows infrastructure providers to manage GPU clusters, while end users only pay for the amount of input and output they generate.

Junghwan Lim, CEO of Korean artificial intelligence startup Motif Technologies, said the company used a $25 million Nvidia B200 GPU, which is backed by the government to develop its latest big language model, Motif 3. Motif 3 scored 47 points in the Artificial Analysis Intelligence Index, which is far above the average of similar models and is on par with Alibaba's Qwen 3.7 Max. Motif is exploring a new token-as-a-service business model that allows infrastructure providers to manage GPU clusters, while end users only pay for the amount of input and output they generate.