After the open-source release of Zhipu GLM-5.1, the domestic large model vendor MiniMax (Xiyu Technology) also officially open-sourced its March 2024 release MiniMax 2.7. However, the "top-tier domestic open-source model" originally highly anticipated has sparked intense debates within the AI community due to significant adjustments in the licensing terms.

Open-Source License Locked Down: Commercial Use Requires Written Permission

Unlike previous relatively lenient open-source licenses, MiniMax 2.7 clearly states: It strictly prohibits any form of commercial use unless it receives written permission from Xiyu Technology. This change directly blocks third-party commercial paths:

  • Aggregation API Platforms: Commercial aggregation API websites like OpenRouter can no longer directly deploy and sell the MiniMax 2.7 interface without authorization.

  • Cloud Service Providers: Domestic cloud service providers such as Alibaba Cloud and Tencent Cloud must first negotiate licensing fees if they want to offer paid services of this model on their platforms.

Performance Close to Opus, but High Barriers for Personal Deployment

According to official data, MiniMax 2.7 scored 56.22% in the programming benchmark test SWE-Pro, with performance extremely close to Claude Opus.

The focus of the community debate lies in the fact that due to the high hardware computing power requirements of this model, ordinary developers and individual users almost cannot achieve local deployment. After the prohibition of commercial platform hosting services, the "open source" of this model is considered to have lost practical significance, more like a publicly available code that is "visible but not usable."

Industry analysts believe that MiniMax's tightening of permissions is largely to prevent other companies from using its free open-source technology for "shell-based" revenue generation. Previously, the foreign AI coding tool Cursor was rumored to have used Kimi 2.5, an open-source large model, to develop its own paid product (later clarified to have obtained authorization). Such incidents have made domestic manufacturers more cautious about "purely free open-source" models.