Model reference · open weights
Minimax-H3-Convrot is an open-weight video model from Abiray, listed in the AxForge catalogue. AxForge can bring it up on EU-owned hardware for you on request — with the licence handled where one is required.
About
MiniMax H3 - Quantized & Pruned Collection (INT4 / INT8 / NVFP4) This repository is a community-compiled collection of quantized and pruned weights for MiniMax H3 (Hailuo 3.0), optimized for local inference environments like ComfyUI. By unifying various quantization formats (INT4, INT8, Mixed, and NVFP4) into a single structured repository, this hub makes it easier for users with consumer GPUs (16GB - 24GB VRAM) to experiment with MiniMax H3's powerful omni-modal text/image/audio-to-video generation capabilities. 💻 Hardware & GPU Selection Guide (Which file to download?) If you are new to local generation and aren't sure what to download, use this guide based on your graphics card. ⚠️ Note on VAEs: Regardless of your GPU, everyone must download both files in the /vae folder (minimaxh3audiovaefp32.safetensors and minimaxh3videovaefp16.safetensors). 1. Standard High-End GPUs (16GB VRAM) Perfect for RTX 4070 Ti Super, RTX 4080, etc. Diffusion Model: Download the INT4 (MiniMaxH3[...]prunedint4convrot.safetensors) or the MIXED (mixedint4int8) version. Text Encoder: Download qwen3vl32bminimaxh3int4convrot.safetensors 2. Enthusiast GPUs (24GB VRAM) Perfect for RTX 3090, RTX 4090, etc. Diffusion Model: Download the INT8 (MiniMaxH3[...]prunedint8convrot.safetensors) for the highest pruned quality. Text Encoder: Download qwen3vl32bminimaxh3int8convrot.safetensors 3. Nvidia Blackwell GPUs ONLY Exclusively for RTX 5090, PRO 6000, and other next-gen Blackwell cards. Do not download these for 30/40 series cards! Diffusion Model: Download the NVFP4 (MiniMaxH3[...]prunednvfp4.safetensors). Text Encoder: Download qwen3vl32bminimaxh3nvfp4awq.safetensors FL2VA vs. Ref2VA (What's the difference?) FL2VA: Download this if you just want to do standard Text-to-Video or simple Image-to-Video (using one or two starting/ending images). Ref2VA: Download this if you want to use advanced omni-references (multiple images, video clips, or audio inputs). 📦 Included Files & Formats This repository uses a unified naming convention (MiniMaxH3FL2VA and MiniMaxH3Ref2VA) for easy integration into your local pipelines. 1. Diffusion Models MIXED INT4/INT8 : mixed-precision models. Requires 15+ VRAM (~1
Summarised from the published model card. Read the full card on the HuggingFace links below.
Specifications
| Maker | Abiray |
|---|---|
| Type | Video models |
| Variants | 1 |
| Runs with | diffusers |
| Based on | MiniMaxAI/MiniMax-H3 |
| Released | 2026-08-03 |
| Popularity | 791k downloads / month |
| Likes | 215 |
| Licence | Commercial licence needed |
How it works
Variants
Open weights ship in several sizes and precisions. One page, all the variants — pick the one that fits your GPU. VRAM figures are estimates from model size.
| Variant | Params | Precision | VRAM | Fits 16 GB | Weights |
|---|---|---|---|---|---|
| Minimax-H3-nvfp4-INT4-INT8-Convrot | — | NVFP4 | — | — | Weights ↗ |
Using it via the API
Once AxForge deploys minimax-h3-convrot for you, it answers on the OpenAI-compatible API — the same base URL and keys as every other model. (minimax-h3-convrot below is illustrative; you get the exact model name on deployment.)
$ curl -sS https://api.axforge.ai/v1/videos/generations \
-H "Authorization: Bearer $AXFORGE_API_KEY" \
-H "Content-Type: application/json" \
-d '{"model":"minimax-h3-convrot","prompt":"a drone shot over a forest"}'
Details
Languages
Tags
Licence
The weights are open but its licence needs a commercial agreement for business use. AxForge can arrange that licence and host the model for you — you pay AxForge, we settle with the model’s maker. Ask us for a quote. Read the licence ↗
Explore