Model reference · open weights
MiniMax-H3-x-Z-Image is an open-weight video model from joeygambino, 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 × Z-Image — the spatial detail graft Z-Image's eye for texture on MiniMax-H3's engine. These are drop-in replacements for the standard MiniMax-H3 checkpoints: same identity, same voices, same speed, same VRAM, same workflows — but sets and surfaces render noticeably richer. Peeling paint peels harder, rust bleeds further, water carries more light. Measured on chained scenes, the extra detail stays flat across joins (no per-shot sharpening creep: 0.99 high-band ratio over 3 joins vs 1.11 on stock). What it looks like (demo videos — three-shot continuous takes rendered with the files on this page) How it works Z-Image (Lumina2, a 6B image model with exceptional texture rendering) and MiniMax-H3 both use per-head Q normalisation in attention. This graft transplants the magnitude profile of Z-Image's spatial attention — how sharply it commits to fine texture — onto H3's later blocks by rescaling qnorm weights. No retraining, no new knowledge, no architecture change: H3 keeps everything it knows and attends to texture the way Z-Image does. Early blocks are left untouched (grafting them produces a lattice artifact in regular textures — measured, not guessed), K normalisation and feed-forward are never touched. This is the second marriage in this line: Joy-LTX 2.5 put JoyAI-Echo's performance on LTX-2.5's engine by weight-delta transplant. Here the donor is a different architecture entirely, so what crosses is attention statistics rather than weights - a mechanism of our own, block-gated and dose-controlled, verified against same-seed baselines. A nod to TenStrip, whose H3 attention experiments sparked the question of what an image model could donate. Which file Same picking rules as standard H3 GGUFs. curve = the current recommended bakes. fl2va vs ref2va: identical choice as stock H3 — ref2va when identity/voice must persist (reference images, voice anchoring, identity bank), fl2va when a shot must land on a supplied frame. Both chain. Install Drop the file where your H3 checkpoints live, pick it in your loader. That's the whole install — every H3 workflow works unchanged, including the MiniMax-H3 Multishot seamless-chain canvases. comfy-native versions (b
Summarised from the published model card. Read the full card on the HuggingFace links below.
Specifications
| Maker | joeygambino |
|---|---|
| Type | Video models |
| Variants | 1 |
| Runs with | minimax-h3 |
| Based on | MiniMaxAI/MiniMax-H3 |
| Released | 2026-08-22 |
| Popularity | 8k downloads / month |
| Likes | 22 |
| 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-x-Z-Image-GGUF | — | GGUF | — | — | Weights ↗ |
Using it via the API
Once AxForge deploys minimax-h3-x-z-image for you, it answers on the OpenAI-compatible API — the same base URL and keys as every other model. (minimax-h3-x-z-image 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-x-z-image","prompt":"a drone shot over a forest"}'
Details
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