MODEL · GPU GUIDE

MiniMax-M3 GPU Requirements: VRAM & Cheapest GPU

MiniMax-M3 has about 427B parameters. See exactly how much GPU memory it needs at FP16, INT8, and INT4, and the cheapest GPU to run it, with live hourly pricing from 5+ data center partners.

427BParameters
233 GBMin VRAM
$3.28/hrCheapest
< 2 minDeploy
MiniMaxAI/MiniMax-M3
VIEW ON HUGGINGFACE ↗
427B paramsimage-text-to-textminimax_m3_vl442 downloads88 likesupdated Jun 12, 2026

To run MiniMax-M3 for inference at FP16, you need roughly 931 GB of VRAM. The cheapest fit on Spheron is 8x H200 141GB at about $14.16/hr. Quantize to INT4 to run it on a smaller, cheaper GPU.

GB VRAM REQUIRED
FP16INFERENCEBATCH 1CTX 4k

Estimated peak VRAM including weights, activations, and KV cache. Add 10% headroom for production traffic.

RANKCONFIGURATIONPER GPUTOTAL $/HR
  • 01
    8× H200 141GBCHEAPEST
    Hopper · HBM3e
    $1.77/hr$14.16/hr
  • 02
    8× B200 192GB
    Blackwell · HBM3e
    $2.71/hr$21.68/hr
  • 03
    4× B300 288GB
    Blackwell Ultra · HBM3e
    $9.02/hr$36.08/hr

Live pricing aggregated from 5+ data center partners. Per-minute billing, no commitments.

VRAM required to run MiniMax-M3

Estimated peak VRAM at context length 4,096 and batch size 1, including weights, activations, and KV cache. Quantizing to INT8 (Q8) or INT4 (Q4) cuts memory roughly in half and in quarter.

PrecisionInferenceLoRA fine-tuneFull fine-tune
FP16931 GB1396 GB3724 GB
INT8465 GB698 GB1862 GB
INT4233 GB349 GB931 GB

Cheapest GPU to run MiniMax-M3 by precision

FP16
VRAM required931GB

Full precision. Best quality, highest memory.

Cheapest GPU
8x H200 141GB
Hopper · HBM3e
$14.16/hr · $1.77/hr/gpu
8x H200 141GB on Spheron
INT8
VRAM required465GB

8-bit quantized. ~2x smaller, minimal quality loss.

Cheapest GPU
8x A100 80GB
Ampere · HBM2e
$6.56/hr · $0.82/hr/gpu
8x A100 80GB on Spheron
INT4
VRAM required233GB

4-bit quantized. ~4x smaller, runs on smaller GPUs.

Cheapest GPU
4x A100 80GB
Ampere · HBM2e
$3.28/hr · $0.82/hr/gpu
4x A100 80GB on Spheron

Inference vs fine-tuning MiniMax-M3

InferenceWeights + KV cache
LoRA fine-tune~1.5×+ low-rank adapter
Full fine-tune~4×+ gradients + optimizer state

Inference only holds the model weights plus a KV cache, so it is the cheapest setup. LoRA fine-tuning adds a small adapter and roughly 50% more memory. Full fine-tuning holds gradients and optimizer state on top of the weights, which is about 4x the inference footprint, so it often needs multiple GPUs even when inference fits on one. For MiniMax-M3, an on-demand H200 141GB instance covers inference and LoRA, while a full fine-tune needs several times that memory and often spans multiple GPUs. Check the live GPU pricing for current rates.

Similar models

Compare GPU requirements for models in the same class.

FAQ / 05

MiniMax-M3 GPU questions