Deploy Long-Context Reasoning and Agentic Workflows with MiniMax M3 on NVIDIA Accelerated Infrastructure | NVIDIA Technical Blog
… For more information, see the vLLM Recipe . vllm serve MiniMaxAI/MiniMax-M3 \ --tensor-parallel-size 8 \ --enable-expert-parallel \ --block-size 128 \ --mm-encoder-attn-backend FLASHINFER \ --mm-processor-cache-type shm \ --tool-call-parser minimax m3 \ --enable-auto-tool-choice \ --reasoning-parse… …