Rent a GPU

MiniMax M3 428B-A23B hardware requirements

427B parameters (23B active, mixture of experts). Needs about 260.5 GB at Q4_K_M with 8K context.

Parameters427B23B active (MoE)
Q4_K_M · 8K260.5 GBthe usual download
FP16 · 8K855.6 GBfull precision
Runs well on0/51GPUs and Macs we track

Memory needed at each quantization

Weights, KV cache and runtime overhead, at 8K and 32K tokens of context.

QuantQuality8K context32K context
FP16Full quality855.6 GB858.6 GB
Q8_0Near-lossless455.3 GB458.3 GB
Q6_KExcellent351.7 GB354.8 GB
Q5_K_MVery good305.3 GB308.3 GB
Q4_K_MGood, most popular260.5 GB263.5 GB
Q3_K_MNoticeably degraded210.3 GB213.3 GB
Q2_KHeavily degraded161.7 GB164.8 GB

Model card

MiniMaxAI/MiniMax-M3 on Hugging Face: licence, prompt format and GGUF builds.

Run it in the cloud

Run MiniMax M3 428B-A23B on a rented GPU today.

It needs about 260.5 GB at Q4_K_M: a rented 4× A100 / H100 runs it for about $4.44 to $15.96 an hour. Compare with Claude and APIs or see how it scores against them.

RunPodRent a GPU on RunPodVast.aiRent a GPU on Vast.ai

Every GPU and Mac checked

Sorted by memory. Bars show how much of each card MiniMax M3 428B-A23B fills at the best quant that fits.

1 Fits with heavy quantization

NameMemory usedBest quantSpeed

50 Too big

NameMemory usedBest quantSpeed