Recently, MiniMax officially announced the open source of its new M2.1 programming model, marking a significant advancement in the field of artificial intelligence. This model is now available on multiple platforms such as Hugging Face, ModelScope, and GitHub, making it easy for developers to quickly access and use it.

The release of M2.1 is a major benefit for developers, as it has received "Day-0" support from vLLM, which means developers can enjoy efficient inference performance right from the moment the model is released. In addition, MiniMax enables efficient FP8 local inference on high-end devices like the latest RTX5090 through KTransformers technology, greatly improving the model's performance and user experience.

image.png

To celebrate this milestone release, MiniMax also launched a series of special promotions designed for developers. Among them, the highly anticipated "9.9 yuan monthly Starter Plan" event has already started on December 26, 2025, and will continue until January 15, 2026, offering great value.

This plan comes in two versions: domestic and international. Domestic users need only pay 9.9 RMB to get the right to use 40 prompts every 5 hours; while international users can get 100 prompts every 5 hours for $2, providing excellent value for money.

In addition, MiniMax is running a "Friend Invitation Event" from December 26, 2025, to February 28, 2026. Participants can enjoy discounts by inviting friends to join MiniMax's Plan, with both parties benefiting. This activity allows users to get a 10% discount when subscribing or upgrading, and it can be combined with the existing Starter Plan first-purchase promotion, further lowering the entry barrier for developers.

With the release of the M2.1 programming model, MiniMax not only provides developers with more powerful tools but also lays the foundation for the implementation and popularization of AI applications. Whether in research fields or commercial applications, M2.1 will help achieve more innovations, and it is worth the attention and trial of a wide range of developers.