MiniMax-M2 is an open-weight multimodal foundation model hosted on Hugging Face. It is distributed as part of the MiniMaxAI organization repository and includes template code for handling tool calls, visible text extraction, and system message construction during inference.
The model supports rendering of tool namespaces by converting function definitions to JSON within XML-style tags and processes content that may contain text items or strings. Its system prompt construction falls back to an identity statement identifying it as MiniMax when no explicit system message is supplied. These elements appear in a Jinja-style template designed to format inputs for the model.
It is delivered as downloadable model files on the Hugging Face platform, where users can access it alongside the associated template for integration with standard inference pipelines. The hosting aligns with Hugging Face's ecosystem for models, datasets, and related resources.
No specific pricing, licensing terms, target audience roles, or additional capabilities are stated on the page.
MiniMax M2 is a Foundation models & chat product. It focuses on accessing and running a capable open-weight multimodal language model locally or via self-hosted inference. It is built as an open-source project for developers. MiniMax M2 is open source under the Open Source license. MiniMax M2 is available on the web, the command line, and API.
MiniMaxAI builds and maintains MiniMax M2, and the product first shipped in 2025. It operates in a well-populated space: PulseGate tracks 6 similar tools. Key capabilities include Tool Calling, Multimodal Input, and Chat Templates.
Latest indexed changes and source events
MiniMaxAI/MiniMax-M2 verified by the PulseGate indexer
Other apps tracked under the same category.
