You can deploy this model and start earning money today!

lukealonso/
MiniMax-M2.5-NVFP4

Name: MiniMax-M2.5-NVFP4
Availability: InStock

safetensors

minimax_m2

custom_code

MiniMaxAI/MiniMax-M2.5

Views

No views yet

Updated 2/16/2026 - Improved calibration coverage.

Model Description

MiniMax-M2.5-NVFP4 is an NVFP4-quantized version of MiniMaxAI/MiniMax-M2.5, a 230B-parameter Mixture-of-Experts language model with 10B active parameters.

The original model weights were converted from the official FP8 checkpoint to BF16, then quantized to NVFP4 (4-bit with blockwise FP8 scales per 16 elements) using NVIDIA Model Optimizer.

What's quantized

Only the MoE expert MLP layers (gate, up, and down projections) are quantized to NVFP4. Attention layers are left in BF16. Since the expert weights constitute the vast majority of model parameters in an MoE architecture, this still yields significant memory savings.

Calibration uses natural top-k routing rather than forcing all experts to activate, so each expert's quantization scales reflect the token distributions it actually sees during inference. To compensate, calibration was run on a vastly larger number of samples than typical to ensure broad expert coverage through natural routing alone.

Calibration dataset

Samples were drawn from a diverse mix of publicly available datasets spanning code generation, function/tool calling, multi-turn reasoning, math, and multilingual (English + Chinese) instruction following. System prompts were randomly varied across samples. The dataset was designed to broadly exercise the model's capabilities and activate diverse token distributions across expert modules.

Quality

MMLU-Pro results (thanks to Lavd for providing these):

Category	Correct	Total	Accuracy
Math	1279	1351	94.7%
Biology	675	717	94.1%
Physics	1188	1299	91.5%
Chemistry	1035	1132	91.4%
Business	715	789	90.6%
Computer Science	366	410	89.3%
Economics	748	844	88.6%
Psychology	674	798	84.5%
Health	686	818	83.9%
Other	767	924	83.0%
Engineering	790	969	81.5%
Philosophy	395	499	79.2%
History	279	381	73.2%
Law	778	1101	70.7%
Overall	10375	12032	86.2%

You should always evaluate against your specific use case.

How to Run

If you experience NCCL hangs with P2P, make sure you have iommu=pt (and amd_iommu=pt on AMD platforms) in your kernel command line.

SGLang

Tested on 2x and 4x RTX Pro 6000 Blackwell.

export NCCL_IB_DISABLE=1
export NCCL_P2P_LEVEL=PHB
export NCCL_ALLOC_P2P_NET_LL_BUFFERS=1
export NCCL_MIN_NCHANNELS=8
export OMP_NUM_THREADS=8
export SAFETENSORS_FAST_GPU=1

python3 -m sglang.launch_server \
  --model lukealonso/MiniMax-M2.5-NVFP4 \
  --served-model-name MiniMax-M2.5 \
  --reasoning-parser minimax \
  --tool-call-parser minimax-m2 \
  --enable-torch-compile \
  --trust-remote-code \
  --tp 2 \
  --mem-fraction-static 0.9 \
  --max-running-requests 16 \
  --kv-cache-dtype bf16 # or fp8_e4m3 \
  --quantization modelopt_fp4 \
  --attention-backend flashinfer \
  --moe-runner-backend flashinfer_cutlass \
  --disable-custom-all-reduce \
  --enable-flashinfer-allreduce-fusion \
  --host 0.0.0.0 \
  --port 8000

vLLM

(thanks to @zenmagnets)

Set your Hugging Face cache and GPUs, then run (from project root with venv activated).

export CUDA_DEVICE_ORDER=PCI_BUS_ID
export CUDA_VISIBLE_DEVICES=0,1
export HF_HOME=/path/to/huggingface
export HUGGINGFACE_HUB_CACHE=$HF_HOME/hub
export VLLM_WORKER_MULTIPROC_METHOD=spawn
export SAFETENSORS_FAST_GPU=1
export VLLM_NVFP4_GEMM_BACKEND=cutlass
export VLLM_USE_FLASHINFER_MOE_FP4=0
export NCCL_IB_DISABLE=1
export OMP_NUM_THREADS=8
export VLLM_ALLOW_LONG_MAX_MODEL_LEN=1

python -m vllm.entrypoints.openai.api_server \
  --model lukealonso/MiniMax-M2.5-NVFP4 \
  --download-dir $HUGGINGFACE_HUB_CACHE \
  --host 0.0.0.0 \
  --port 1235 \
  --served-model-name MiniMax-M2.5-NVFP4 \
  --trust-remote-code \
  --tensor-parallel-size 2 \
  --attention-backend FLASH_ATTN \
  --gpu-memory-utilization 0.95 \
  --max-model-len 190000 \
  --max-num-batched-tokens 16384 \
  --max-num-seqs 64 \
  --disable-custom-all-reduce \
  --enable-auto-tool-choice \
  --tool-call-parser minimax_m2 \
  --reasoning-parser minimax_m2_append_think

Dependencies

Install in a Python 3.12 venv; use CUDA 12.x on the host.

Package	Version	Note
vllm	0.15.1	OpenAI server + NVFP4 MoE
torch	2.9.1+cu128	CUDA 12.8 build
transformers	4.57.6	
safetensors	0.7.0	
nvidia-modelopt	0.41.0	NVFP4 / ModelOpt format
flashinfer-python	0.6.1	Optional (we use FLASH_ATTN)
nvidia-nccl-cu12	2.27.5	Multi-GPU
nvidia-cutlass-dsl*	4.4.0.dev1	NVFP4 GEMM (script uses cutlass backend)
System: CUDA 12.8, cuDNN 9.10.2 (or matching torch cuDNN). Driver must support your GPUs (e.g. Blackwell).