{ "hf_id": "MiniMaxAI/MiniMax-M2.5", "meta": { "title": "MiniMax-M2.5", "slug": "minimax-m2.5", "provider": "MiniMax", "description": "MiniMax M2.5 MoE language model (230B total / 10B active) for coding, agent toolchains, and long-context reasoning — native FP8 checkpoint", "date_added": "2026-02-13", "date_updated": "2026-09-14", "tasks": [ "text" ], "performance_headline": "Refreshed M2 series MoE with strong SWE-Bench and Terminal-Bench performance, 196K context", "related_recipes": [], "hardware": { "h100": "verified", "h200": "verified", "mi300x": "verified", "mi325x": "verified", "mi355x": "verified" } }, "recommended_command": { "hardware": "h200", "strategy": "single_node_tp", "variant": "default", "node_count": 1, "deploy_type": "single_node", "env": { "PYTHONNOUSERSITE": "1", "VLLM_USE_DEEP_GEMM": "0", "VLLM_BLOCKSCALE_FP8_GEMM_FLASHINFER": "0", "VLLM_FLOAT32_MATMUL_PRECISION": "high", "VLLM_USE_RUST_FRONTEND": "1" }, "docker_image": "vllm/vllm-openai:v0.28.0", "command": "vllm serve MiniMaxAI/MiniMax-M2.5 \\\n --trust-remote-code \\\n --tensor-parallel-size 4 \\\n --compilation-config '{\"cudagraph_mode\":\"PIECEWISE\"}' \\\n --max-num-seqs 512 \\\n --max-num-batched-tokens 32768 \\\n --kv-cache-dtype fp8 \\\n --moe-backend triton \\\n --attention-backend FLASHINFER \\\n --enable-flashinfer-autotune \\\n --tool-call-parser minimax_m2 \\\n --enable-auto-tool-choice \\\n --reasoning-parser minimax_m2", "argv": [ "vllm", "serve", "MiniMaxAI/MiniMax-M2.5", "--trust-remote-code", "--tensor-parallel-size", "4", "--compilation-config", "{\"cudagraph_mode\":\"PIECEWISE\"}", "--max-num-seqs", "512", "--max-num-batched-tokens", "32768", "--kv-cache-dtype", "fp8", "--moe-backend", "triton", "--attention-backend", "FLASHINFER", "--enable-flashinfer-autotune", "--tool-call-parser", "minimax_m2", "--enable-auto-tool-choice", "--reasoning-parser", "minimax_m2" ], "docker_command": "docker run --gpus all \\\n --privileged --ipc=host -p 8000:8000 \\\n -v ~/.cache/huggingface:/root/.cache/huggingface \\\n -e PYTHONNOUSERSITE=1 \\\n -e VLLM_USE_DEEP_GEMM=0 \\\n -e VLLM_BLOCKSCALE_FP8_GEMM_FLASHINFER=0 \\\n -e VLLM_FLOAT32_MATMUL_PRECISION=high \\\n -e VLLM_USE_RUST_FRONTEND=1 \\\n vllm/vllm-openai:v0.28.0 MiniMaxAI/MiniMax-M2.5 \\\n --trust-remote-code \\\n --tensor-parallel-size 4 \\\n --compilation-config '{\"cudagraph_mode\":\"PIECEWISE\"}' \\\n --max-num-seqs 512 \\\n --max-num-batched-tokens 32768 \\\n --kv-cache-dtype fp8 \\\n --moe-backend triton \\\n --attention-backend FLASHINFER \\\n --enable-flashinfer-autotune \\\n --tool-call-parser minimax_m2 \\\n --enable-auto-tool-choice \\\n --reasoning-parser minimax_m2", "docker_argv": [ "docker", "run", "--gpus", "all", "--privileged", "--ipc=host", "-p", "8000:8000", "-v", "~/.cache/huggingface:/root/.cache/huggingface", "-e", "PYTHONNOUSERSITE=1", "-e", "VLLM_USE_DEEP_GEMM=0", "-e", "VLLM_BLOCKSCALE_FP8_GEMM_FLASHINFER=0", "-e", "VLLM_FLOAT32_MATMUL_PRECISION=high", "-e", "VLLM_USE_RUST_FRONTEND=1", "vllm/vllm-openai:v0.28.0", "MiniMaxAI/MiniMax-M2.5", "--trust-remote-code", "--tensor-parallel-size", "4", "--compilation-config", "{\"cudagraph_mode\":\"PIECEWISE\"}", "--max-num-seqs", "512", "--max-num-batched-tokens", "32768", "--kv-cache-dtype", "fp8", "--moe-backend", "triton", "--attention-backend", "FLASHINFER", "--enable-flashinfer-autotune", "--tool-call-parser", "minimax_m2", "--enable-auto-tool-choice", "--reasoning-parser", "minimax_m2" ], "strategy_spec": { "name": "single_node_tp", "deploy_type": "single_node", "display_name": "Tensor Parallel", "short_label": "TP", "orientation": "latency", "description": "Single-node tensor parallel. Splits the model across all local GPUs. TP size is set to the GPU count at deploy time. The simplest multi-GPU strategy — works for all model architectures.\n", "hardware_match": { "min_gpus": 1, "max_gpus": 8, "multi_node": false }, "vllm_args": [], "parallel_flag": "--tensor-parallel-size" }, "hardware_profile": { "brand": "NVIDIA", "generation": "hopper", "compute_arch": "sm90", "display_name": "H200", "description": "NVIDIA H200 SXM 141 GB HBM3e · 8-GPU HGX node", "gpu_count": 8, "vram_gb": 1128, "multi_node": false }, "alternatives": { "single_node_tep": "/MiniMaxAI/MiniMax-M2.5/strategies/single_node_tep.json", "single_node_dep": "/MiniMaxAI/MiniMax-M2.5/strategies/single_node_dep.json", "multi_node_tp": "/MiniMaxAI/MiniMax-M2.5/strategies/multi_node_tp.json", "multi_node_tp_pp": "/MiniMaxAI/MiniMax-M2.5/strategies/multi_node_tp_pp.json", "multi_node_tep": "/MiniMaxAI/MiniMax-M2.5/strategies/multi_node_tep.json", "multi_node_dep": "/MiniMaxAI/MiniMax-M2.5/strategies/multi_node_dep.json", "pd_cluster": "/MiniMaxAI/MiniMax-M2.5/strategies/pd_cluster.json", "kv_store_centralized_mooncake": "/MiniMaxAI/MiniMax-M2.5/strategies/kv_store_centralized_mooncake.json", "kv_store_distributed_mooncake": "/MiniMaxAI/MiniMax-M2.5/strategies/kv_store_distributed_mooncake.json" }, "by_hardware": { "h200": "/MiniMaxAI/MiniMax-M2.5/hw/h200.json", "h100": "/MiniMaxAI/MiniMax-M2.5/hw/h100.json", "b200": "/MiniMaxAI/MiniMax-M2.5/hw/b200.json", "gb200": "/MiniMaxAI/MiniMax-M2.5/hw/gb200.json", "b300": "/MiniMaxAI/MiniMax-M2.5/hw/b300.json", "gb300": "/MiniMaxAI/MiniMax-M2.5/hw/gb300.json", "mi300x": "/MiniMaxAI/MiniMax-M2.5/hw/mi300x.json", "mi325x": "/MiniMaxAI/MiniMax-M2.5/hw/mi325x.json", "mi350x": "/MiniMaxAI/MiniMax-M2.5/hw/mi350x.json", "mi355x": "/MiniMaxAI/MiniMax-M2.5/hw/mi355x.json" } }, "model": { "model_id": "MiniMaxAI/MiniMax-M2.5", "min_vllm_version": "0.20.2", "default_frontend": "rust", "docker_image": "vllm/vllm-openai:v0.28.0", "architecture": "moe", "parameter_count": "230B", "active_parameters": "10B", "context_length": 196608, "base_args": [ "--trust-remote-code" ], "base_env": {}, "install": { "pip": { "command": "uv venv\nsource .venv/bin/activate\nuv pip install -U vllm --torch-backend auto" }, "docker": { "command": "docker pull vllm/vllm-openai:v0.28.0" } } }, "features": { "tool_calling": { "description": "MiniMax M2 tool call parser with automatic tool choice", "args": [ "--tool-call-parser", "minimax_m2", "--enable-auto-tool-choice" ] }, "reasoning": { "description": "MiniMax M2 reasoning parser for chain-of-thought extraction", "args": [ "--reasoning-parser", "minimax_m2" ] } }, "opt_in_features": [], "variants": { "default": { "precision": "fp8", "vram_minimum_gb": 276, "description": "Native FP8 checkpoint — 4x H200/H20/H100 or 4x A100/A800 for weights, plus KV cache headroom" }, "nvfp4": { "model_id": "nvidia/MiniMax-M2.5-NVFP4", "precision": "nvfp4", "vram_minimum_gb": 138, "description": "NVIDIA NVFP4 quantized weights for Blackwell GPUs", "extra_args": [ "--kv-cache-dtype", "fp8" ], "json": "/nvidia/MiniMax-M2.5-NVFP4.json" }, "redhat_nvfp4": { "model_id": "RedHatAI/MiniMax-M2.5-NVFP4", "precision": "nvfp4", "label": "NVFP4 (RedHat)", "vram_minimum_gb": 146, "description": "RedHatAI NVFP4 quantization", "json": "/RedHatAI/MiniMax-M2.5-NVFP4.json" } }, "kv_offload_support": { "offloading_cpu": "verified", "offloading_fs": "verified" }, "hardware_overrides": { "nvidia": { "extra_args": [ "--compilation-config", "{\"cudagraph_mode\":\"PIECEWISE\"}" ] }, "hopper": { "extra_args": [ "--compilation-config", "{\"cudagraph_mode\":\"PIECEWISE\"}", "--max-num-seqs", "512", "--max-num-batched-tokens", "32768", "--kv-cache-dtype", "fp8", "--moe-backend", "triton", "--attention-backend", "FLASHINFER", "--enable-flashinfer-autotune" ], "extra_env": { "PYTHONNOUSERSITE": "1", "VLLM_USE_DEEP_GEMM": "0", "VLLM_BLOCKSCALE_FP8_GEMM_FLASHINFER": "0", "VLLM_FLOAT32_MATMUL_PRECISION": "high" } }, "amd": { "extra_args": [ "--tensor-parallel-size", "2", "--attention-backend", "ROCM_AITER_FA" ], "extra_env": { "VLLM_ROCM_USE_AITER": "1", "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1" } } }, "guide": "## Overview\n\n[MiniMax-M2.5](https://huggingface.co/MiniMaxAI/MiniMax-M2.5) is part of the MiniMax\nM2 series of advanced MoE language models. It retains the M2 architecture (10B active,\n230B total) and a 196K context per sequence.\n\n## Prerequisites\n\n- **OS:** Linux\n- **Python:** 3.10 - 3.13\n- **NVIDIA:** compute capability >= 7.0; ~220 GB for weights + 240 GB per 1M context tokens\n- **AMD:** MI300X / MI325X / MI350X / MI355X with ROCm 7.0+\n\n### Install vLLM (NVIDIA)\n\n```bash\nuv venv\nsource .venv/bin/activate\nuv pip install -U vllm --torch-backend auto\n```\n\n### Docker (dedicated M2-series image)\n\n```bash\ndocker run --gpus all \\\n -p 8000:8000 \\\n --ipc=host \\\n -v ~/.cache/huggingface:/root/.cache/huggingface \\\n -e PYTHONNOUSERSITE=1 \\\n -e VLLM_USE_DEEP_GEMM=0 \\\n -e VLLM_BLOCKSCALE_FP8_GEMM_FLASHINFER=0 \\\n -e VLLM_FLOAT32_MATMUL_PRECISION=high \\\n vllm/vllm-openai:v0.28.0 MiniMaxAI/MiniMax-M2.5 \\\n --tensor-parallel-size 4 \\\n --max-num-seqs 512 \\\n --max-num-batched-tokens 32768 \\\n --kv-cache-dtype fp8 \\\n --moe-backend triton \\\n --attention-backend FLASHINFER \\\n --enable-flashinfer-autotune \\\n --tool-call-parser minimax_m2 \\\n --reasoning-parser minimax_m2 \\\n --enable-auto-tool-choice \\\n --compilation-config '{\"cudagraph_mode\":\"PIECEWISE\"}' \\\n --trust-remote-code\n```\n\n## Launching the Server\n\n### NVIDIA H200 — TP4\n\n```bash\nPYTHONNOUSERSITE=1 \\\nVLLM_USE_DEEP_GEMM=0 \\\nVLLM_BLOCKSCALE_FP8_GEMM_FLASHINFER=0 \\\nVLLM_FLOAT32_MATMUL_PRECISION=high \\\nvllm serve MiniMaxAI/MiniMax-M2.5 \\\n --tensor-parallel-size 4 \\\n --max-num-seqs 512 \\\n --max-num-batched-tokens 32768 \\\n --kv-cache-dtype fp8 \\\n --moe-backend triton \\\n --attention-backend FLASHINFER \\\n --enable-flashinfer-autotune \\\n --tool-call-parser minimax_m2 \\\n --reasoning-parser minimax_m2 \\\n --compilation-config '{\"cudagraph_mode\":\"PIECEWISE\"}' \\\n --enable-auto-tool-choice \\\n --trust-remote-code\n```\n\nFor H200 FP8, choose the attention backend by sequence length for best performance:\nshorter sequences (e.g. 1024) should keep the command above with `--attention-backend FLASHINFER`\nand `--enable-flashinfer-autotune`, while longer input sequences (e.g. 8192)\nshould remove those flags to use the auto-selected FlashAttention backend.\n\nPure TP8 is not supported. For >4 GPUs use DP+EP or TP+EP.\n\n### TP4+EP (recommended for H100)\n\n```bash\nvllm serve MiniMaxAI/MiniMax-M2.5 \\\n --tensor-parallel-size 4 \\\n --enable-expert-parallel \\\n --tool-call-parser minimax_m2 \\\n --reasoning-parser minimax_m2 \\\n --compilation-config '{\"cudagraph_mode\":\"PIECEWISE\"}' \\\n --enable-auto-tool-choice\n```\n\n### AMD ROCm — TP2\n\n```bash\nVLLM_ROCM_USE_AITER=1 vllm serve MiniMaxAI/MiniMax-M2.5 \\\n --tensor-parallel-size 2 \\\n --tool-call-parser minimax_m2 \\\n --reasoning-parser minimax_m2 \\\n --enable-auto-tool-choice \\\n --trust-remote-code\n```\n\n## Benchmarking\n\n```bash\nvllm bench serve \\\n --backend vllm \\\n --model MiniMaxAI/MiniMax-M2.5 \\\n --endpoint /v1/completions \\\n --dataset-name random \\\n --random-input 2048 \\\n --random-output 1024 \\\n --max-concurrency 10 \\\n --num-prompt 100\n```\n\n## Troubleshooting\n\n- See [MiniMax-M2](./MiniMax-M2.yaml) for shared troubleshooting notes\n (latest vLLM fusion behavior, nightly vs stable, DeepGEMM, AITER).\n\n## References\n\n- [Model card](https://huggingface.co/MiniMaxAI/MiniMax-M2.5)\n- [MiniMax](https://www.minimax.io/)\n", "source": { "source": "vllm", "repo": "vllm-project/vllm", "ref": "main", "sha": "389ffebe555705c85337025d5df108d0384d365e", "path": "recipes/models/MiniMaxAI/MiniMax-M2.5.yaml", "edit_url": "https://github.com/vllm-project/vllm/edit/main/recipes/models/MiniMaxAI/MiniMax-M2.5.yaml", "overlays": [] } }