The "Big Three" American AI giants—OpenAI, Anthropic, and Google—are currently in talks to collaborate in an effort to curb Chinese competitors from extracting results from advanced US AI models (a process known as AI model distillation) to gain a leading edge in the global AI race.
Previously, Chinese companies such as DeepSeek, MiniMax, and Moonshot AI have attempted, to varying degrees, to distill massive amounts of data from these advanced models through bulk account registration, data scraping, and model distillation techniques. This practice makes the resulting distilled AI models significantly cheaper than their "teacher models" (like GPT, Claude, and Gemini), while also delivering faster output speeds and higher efficiency. This has allowed them to secure quite a few advantages in the AI sector.
Especially following the launch of Openclaw, which many people have started deploying, the token consumption required is immense. DeepSeek and MiniMax, with their API costs being significantly lower than those of GPT and Claude, quickly dominated the mainstream market right as the trend took off.
The image above displays data from OpenRouter, showing that MiniMax's M2.5 held the number one spot for API Token model usage for a long time spanning February and March.
This has forced the "Big Three," who are traditionally fierce rivals, to join forces and share data to intensify their crackdown on model distillation behaviors.
{dotted startColor="#ff6c6c" endColor="#1989fa"/}
References / Sources: Bloomberg
Comments NOTHING