Domestic large-scale models have topped the global charts for five consecutive weeks.

Domestic large model MiniMax M2.5 has topped the global large model invocation rankings for five consecutive weeks. At the company’s headquarters office, product development staff showed reporters a clear price difference. “Models overseas with the same capabilities may be ten times more expensive than ours.” Cost performance is becoming the core competitive advantage of domestic large models in attracting global users. Where does this cost performance come from? The first key factor is technological innovation. By innovating at the underlying architecture, it fundamentally reduces inference costs. Simply put, it accomplishes the same tasks with fewer tokens. Another key factor is energy advantage. Industry insiders told reporters that currently, electricity costs account for 70% to 80% of computing power costs. For AI clusters that require millions of GPUs running in parallel, even a few cents difference in electricity prices translates into astronomical annual operating costs. Stable energy supply and relatively low electricity prices have built a more competitive cost defense for China’s AI industry.