MiniMax H3
VideoPopularMiniMax · 33B · Dense
Open video generation — text/image to 2K video with native stereo audio
0
0
2026-08
32K tokens
Use cases
Your hardware
16 GB Reference Desktop
Quantization options
What is quantization?| Quant | Bits | VRAM | Disk | Quality | Status |
|---|---|---|---|---|---|
| Q2_K | 2 | 11.1 GB | 10.1 GB | low | Tight fit · ~18 t/s |
| Q3_K_M | 3 | 15.3 GB | 14.1 GB | moderate | Barely runs · ~13 t/s |
| Q4_K_M | 4 | 17.4 GB | 16.1 GB | good | Too heavy · ~7 t/s |
| Q5_K_M | 5 | 21.6 GB | 20.2 GB | good | Barely runs · ~4 t/s |
| Q6_K | 6 | 25.9 GB | 24.2 GB | excellent | Barely runs · ~2 t/s |
| Q8_0 | 8 | 34.3 GB | 32.3 GB | excellent | Barely runs · ~1 t/s |
| F16 | 16 | 68.1 GB | 64.5 GB | lossless | Too heavy · ~1 t/s |
About this model
MiniMax H3 (MiniMax h3) is a 33B video generation architecture released by MiniMax. Unlike autoregressive text LLMs whose KV cache grows with chat turns, local image and video diffusion pipelines allocate memory simultaneously for the transformer backbone, companion text encoders (such as T5/CLIP), and high-resolution VAE latent decoding buffers.
Baseline sizing plans for 18.4 GB minimum memory and 30.7 GB recommended memory, with the Q4_K_M checkpoint requiring roughly 17.4 GB of VRAM and 16.1 GB of SSD storage. Exceeding dedicated VRAM forces PCIe weight swapping across host RAM and slows step sampling dramatically.
Desktop NVIDIA RTX workstations using ComfyUI or Diffusers with CUDA acceleration deliver the cleanest generation throughput, while high-memory Apple Silicon Macs can execute workloads via MPS with conservative batch size 1 and lower output dimensions. Under the MiniMax Community license, verify commercial deployment terms on the official repository card before production use.
Highlights
- Generation VRAM Budget18.4 GB minimum and 30.7 GB recommended across backbone, encoder, and VAE buffers.
- Baseline QuantizationQ4_K_M targets 17.4 GB memory and 16.1 GB storage.
- Recommended PipelineComfyUI or Diffusers on NVIDIA CUDA desktop GPUs or high-RAM Apple Silicon Macs.
System requirements by device
Memory planning targets 18.4 GB minimum RAM/VRAM and 30.7 GB recommended memory. The baseline quant is Q4_K_M, estimated at 17.4 GB memory and 16.1 GB disk. Actual runtime usage shifts with context length, KV cache precision, and OS background overhead.
| Device | Notes |
|---|---|
| Mac (Apple silicon) | Use unified memory headroom (~75% physical RAM max) as the primary fit signal. Start with Q4_K_M and close heavy browser tabs first. |
| Windows / Linux GPU | Prefer dedicated NVIDIA RTX VRAM via CUDA (ComfyUI / Diffusers) so UNet, text encoder, and VAE buffers stay off PCIe RAM. |
| Phones & Compact Edge Boards | Not supported locally: mobile RAM ceilings and app watchdogs cannot hold multi-stage diffusion and VAE buffers. |
Start with a short session and leave memory for the operating system. A file fitting on disk does not guarantee that it will load in memory.
Estimated memory by quantization
These are rough estimates, not tested minimum requirements. Check the actual file and runtime documentation; context, offloading and operating-system overhead can change the memory needed.
| Quant Tier | Estimated memory | Disk Size | Quality |
|---|---|---|---|
| Q2_K | 11.1 GB | 10.1 GB | low |
| Q3_K_M | 15.3 GB | 14.1 GB | moderate |
| Q4_K_M | 17.4 GB | 16.1 GB | good |
| Q5_K_M | 21.6 GB | 20.2 GB | good |
| Q6_K | 25.9 GB | 24.2 GB | excellent |
| Q8_0 | 34.3 GB | 32.3 GB | excellent |
| F16 | 68.1 GB | 64.5 GB | lossless |
How to run MiniMax H3 on your device
Start with a short prompt and the smallest quant that still gives the quality you need. A clean first run is more useful than loading a larger file with no memory headroom.
Run MiniMax H3 on Windows / Linux GPU
- Install ComfyUI or a current Diffusers Python environment with PyTorch CUDA enabled.
- Select a quantized checkpoint such as Q4_K_M (17.4 GB) and confirm at least 18.4 GB of free VRAM.
- Keep VAE tiling enabled and test a standard resolution (e.g., 1024x1024 or short frame count) before increasing batch or step counts.
- Avoid spilling diffusion UNet/DiT weights across PCIe into system RAM if interactive generation speed matters.
Run MiniMax H3 on Apple Silicon Mac
- Confirm your Mac has at least 30.7 GB unified memory so Metal buffers do not force macOS swap.
- Load Q4_K_M weights inside ComfyUI (MPS) or a native Apple Silicon Diffusers workflow.
- Close memory-heavy browsers and keep latent batch size at 1 during initial generation tests.
- Consult the Mac hardware guide when planning multi-step image or video pipelines.
MiniMax H3 vs alternatives
| Model | Params | Context | Min RAM | License |
|---|---|---|---|---|
| MiniMax H3 | 33B | 32K | 18.4 GB | MiniMax Community |
| MiniMax M3 | 428B (23B active) | 1M | 239.2 GB | MiniMax Community |
| Gemma 4 31B IT | 33B | 256K | 18.4 GB | Gemma |
| Gemma 4 31B | 33B | 256K | 18.4 GB | Gemma |
| Qwen 2.5 32B | 32B | 128K | 17.9 GB | Apache 2.0 |
| Qwen 2.5 Coder 32B | 32B | 128K | 17.9 GB | Apache 2.0 |
Use this table to choose the smallest nearby model that still matches the job. A related model with lower RAM can be the better local choice when the current page fits only at the edge of available memory.
Frequently asked questions
- What is MiniMax H3?
- MiniMax H3 is a 33B media generation model from MiniMax. Sizing estimates allocate 18.4 GB minimum VRAM and 30.7 GB recommended memory for weights, text encoders, and VAE decode buffers.
- How do I download and install MiniMax h3?
- Download a GGUF build for MiniMax H3, start with Q4_K_M, and confirm the model fits locally before increasing context or quant quality.
- How much VRAM or RAM does MiniMax H3 require?
- At Q4_K_M, MiniMax H3 occupies roughly 17.4 GB of active memory and 16.1 GB on disk. Add 2–6 GB of spare headroom for KV cache buffers, runtime scratch space, and OS apps (30.7 GB recommended).
- Which software runs MiniMax H3 locally?
- Run MiniMax H3 inside ComfyUI or Hugging Face Diffusers on an NVIDIA CUDA desktop GPU or high-memory Apple Silicon Mac. Start with Q4_K_M or FP8 builds before testing higher-precision weights.
- Can phones run MiniMax H3 locally?
- No. Desktop image and video diffusion checkpoints require high-bandwidth desktop GPUs or high-memory Apple Silicon Macs rather than smartphone runtimes.
- How does MiniMax H3 compare to MiniMax M3?
- MiniMax M3 (428B, 239.2 GB min RAM) sits closest in size. Compare active parameters, context ceiling (32K vs 1M), and license terms (MiniMax Community) before switching.