Skip to content
English

Switch theme

// MiniMax · Video generation

MiniMax-H3: VRAM & local deployment

33.1B parameters. About 22.6 GB at 4-bit and 73.4 GB at BF16.

released 2026-07-28licence: Custom licence – read itmemory figures are estimates

01Specs

PublisherMiniMax · MiniMaxAI
Family / seriesMiniMax · MiniMax-H3
TypeVideo generation (image-text-to-video)
Parameters33.1B Source: counted from official safetensors weights
Released weightsBF16 · 498 GBdownload
Licenceminimax-h3-community-license-agreement · Custom licence – read itThe publisher uses its own licence; we have not classified it. Read the licence text.
Released2026-07-28 (Hugging Face repo created)

02How much VRAM it needs (estimate)

Video models: memory grows with resolution and clip length; official repos usually list a tested GPU in the card. Weights below are the main model only.

PrecisionWeightsWeights + overheadSmallest GPUMac
BF16 / FP16 (original)66.3 GB 73.4 GB RTX PRO 6000 96GBMac 128GB
FP8 / INT8 / Q8_035.2 GB 39.2 GB 2 × RTX 5090 (64GB)Mac 64GB
4-bit (Q4_K_M / AWQ / GPTQ-Int4)20.1 GB 22.6 GB RTX 5090 32GBMac 64GB

Bits per weight: BF16 16, FP8 8.5, 4-bit 4.85. “Fits” = total ≤ 90% of memory; Mac GPU can use ~2/3 of unified memory (≤36GB) or ~3/4 (larger). For MoE, all experts must be in memory, so size by total parameters. Method →

03How to run it

Only runtimes the publisher or the runtime’s official library names for this model.

Which inference engine to choose →

04Official repositories & quantised builds

1 official repo(s) from MiniMaxAI.

05Other MiniMax models

All 11 →
ModelParams4-bit totalReleased
MiniMax-Music32.43B2.1 GB2026-08-07
MiniMax-M3428B / 23B286 GB2026-06-02
MiniMax-M2.7228.7B155 GB2026-04-09
MiniMax-M2.5228.7B155 GB2026-02-12
MiniMax-M2.1228.7B155 GB2025-12-20
MiniMax-M2230B / 10B156 GB2025-10-22
MiniMax-M1-80k456B / 45.9B305 GB2025-06-13
MiniMax-M1-40k456B / 45.9B305 GB2025-06-05

06Sources

Data checked 2026-08-13. Parameters, context and licence are copied from the official repository; memory figures are our estimates.

Data updated: Self-host open models(UTC+8)