This is an NVIDIA-optimized FP4 quantized version of the MiniMax-M3 multimodal foundation model. It supports text, image, and video understanding along with advanced tool-calling features. The quantization enables faster and more memory-efficient inference on NVIDIA hardware while preserving model capabilities.
MiniMax M3 is a Foundation models & chat product. It focuses on running large multimodal models efficiently with reduced memory and compute requirements. It is built as an open-source project for developers. MiniMax M3 is open source under the Apache-2.0 license. The product ships for the web and API.
NVIDIA builds and maintains MiniMax M3, and the product first shipped in 2024. Development happens publicly on GitHub with 3.3k stars and 356 commits in the last 90 days. Key capabilities include Multimodal Processing, Tool Calling, and FP4 Quantization.
Latest indexed changes and source events
nvidia/MiniMax-M3-NVFP4 verified by the PulseGate indexer
Other apps tracked under the same category.
