On March 2, 2026, Akshay Kothari, co-founder of the globally renowned all-in-one workspace platform Notion, announced that its AI automation tool, Notion Custom Agents, has officially integrated the domestic open-source weight model MiniMax M2.5, and is now offering it as an experimental feature to over 100 million users worldwide.

This move marks the entry of MiniMax M2.5 into Notion's core model library alongside internationally renowned closed-source models such as Claude Sonnet 4.6, Opus 4.6, Haiku 4.5, and GPT-5.2, GPT-5.3 Codex, becoming an important member of the platform's model selection list.

QQ20260302-135554.png

As the only "open-source weight model" in the list, the integration of MiniMax M2.5 provides Notion with a more cost-effective computing power solution for document, note-taking, and project management scenarios. Kothari pointed out that the usage cost of this model is significantly lower than traditional closed-source large models when performing simple tasks. This move by Notion reflects a shift in the model integration strategy of productivity tools, from solely relying on top closed-source models to a hybrid architecture combining "high-performance closed-source" and "cost-effective open-source" models.

This trend not only helps reduce the AI call threshold for enterprise users but also indicates that domestic models are further solidifying their collaborative position within international mainstream ecosystems. As competition in the global productivity tools market intensifies, optimizing task efficiency and operational costs through diversified model configurations has become an inevitable path for industry development.