Access to over 300 models, token cost reduced by 30 percent, more than 100 billion calls per day
Let’s try to understand this in a CCTV+ report.
China Mobile, the country’s telecommunications giant, launched a platform on Friday that brings together the largest number of artificial intelligence models in the country, providing users with access to more than 300 core models and their services in one place.
The MoMA platform can automatically analyze task requirements and intelligently assign the optimal model for a given task from three options: “cost priority”, “performance priority” and “balanced”. This allows tasks to be completed with lower token consumption and higher execution efficiency.
If a model encounters a timeout, rate limit or technical failure, the platform can seamlessly switch to another model within seconds, ensuring uninterrupted service.
Li Li, head of intelligent computing products at the cloud capabilities center of China Mobile Communication Corporation, said: “For a simple task, we can use a low-cost model that consumes very few tokens. For a complex one, we will turn to a more intelligent model that consumes more tokens but is very smart. And throughout this switching process, the platform never compromises on accuracy.”
The platform has currently reduced the cost of a single token by approximately 30 percent and processes an average of more than 100 billion calls per day.
According to the National Data Administration, token usage is showing exponential growth this year as AI applications accelerate their real-world deployment. As of the end of March, the average daily volume exceeded 140 trillion — more than a thousand times higher than at the end of 2024.
The answer is found: it’s not magic and not one supermodel. It’s a smart dispatch system that knows who, when and at what price to send to a task.