The nvidia/MiniMax-M2.7-NVFP4 is a quantized version of the MiniMax-M2.7 language model hosted on Hugging Face. It is published by NVIDIA and uses the NVFP4 quantization format.
The model page supplies a set of Jinja2 rendering macros and templates that define how system messages, tool calls, and visible text content are constructed for structured interactions. These include a render_tool_namespace macro that outputs tool definitions in XML-style tags, a visible_text macro that handles string or iterable content with text-type items, and a build_system_message macro that extracts content from a system message object or defaults to an identity string naming the model MiniMax-M2.7. The templates also specify token delimiters such as <minimax:tool_call> and </minimax:tool_call>.
It is delivered as a model repository on the Hugging Face platform, where it can be accessed alongside other models, datasets, and spaces. The page forms part of the broader Hugging Face ecosystem that includes documentation, community forums, and enterprise offerings.
In the Foundation models & chat space, MiniMax M takes a focused approach. It focuses on delivering an NVIDIA-optimized quantized version of the MiniMax M2.7 model for efficient local inference. It is built as an open-source project for developers. MiniMax M is open source under the Open Source license. MiniMax M is available on the web and API.
Behind MiniMax M is NVIDIA, and the product first shipped in 2026. Development happens publicly on GitHub with 354 stars. Key capabilities include Tool Calling, Quantized Model, and System Prompts.
Latest indexed changes and source events
nvidia/MiniMax-M2.7-NVFP4 verified by the PulseGate indexer
Other apps tracked under the same category.
