Views
No views yet
10Eros_Max_h3_ref2va_beta2_pruned version).ComfyUI-GGUF, allowing users with consumer GPUs (12GB - 16GB VRAM) to run this massive 40GB+ MiniMax-H3 model locally. The ref2va version features TenStrip's specific grafting recipe combining weights from LTX, Wan, and Krea.| File | Size | Description | VRAM Target |
|---|---|---|---|
10Eros-Max-Ref2VA-Beta2-Pruned-Q3_K_M.gguf | 8.9 GB | Extreme compression. Expect some degradation in fine details. | <12GB |
10Eros-Max-Ref2VA-Beta2-Pruned-Q4_K_M.gguf | 11.6 GB | ⭐️ Recommended. Perfect balance of quality and size. Preserves text and physics incredibly well. | 12GB - 16GB |
10Eros-Max-Ref2VA-Beta2-Pruned-Q4_K_S.gguf | 11.6 GB | Slightly smaller alternative to Q4_K_M. | 12GB - 16GB |
10Eros-Max-Ref2VA-Beta2-Pruned-Q5_K_M.gguf | 14.1 GB | High fidelity. Great for complex geometry and micro-details. | 16GB - 24GB |
10Eros-Max-Ref2VA-Beta2-Pruned-Q5_K_S.gguf | 14.1 GB | Slightly smaller alternative to Q5_K_M. | 16GB - 24GB |
10Eros-Max-Ref2VA-Beta2-Pruned-Q6_K.gguf | 16.7 GB | Near-lossless visual quality. | 24GB+ |
10Eros-Max-Ref2VA-Beta2-Pruned-Q8_0.gguf | 21.6 GB | Maximum quality, nearly indistinguishable from original BF16. | 24GB+ |
ComfyUI-GGUF (by city96) via the ComfyUI Manager..gguf file (Q4_K_M is recommended)..gguf file inside your ComfyUI/models/unet directory.Load Diffusion Model node with the Unet Loader (GGUF) node.euler / simple or flowmatch)For TURBO model use: multires/simple 6 steps, er_sde/simple 6 steps, or I prefer and merged it for er_sde with custom 7 step sigmas to get rid of motion noise details. Er_sde/simple and custom sigma string instead of basic scheduler:[1.00, 0.94, 0.83, 0.72, 0.55, 0.30, 0.10, 0.00]
Turbo fromsilveroxides/MiniMax-H3_tests/tree/mainwith silu(t_emb) grid restored before merge.
Due to a ton of confusion I've let Claude compile a full MD on the grafting, including code and methodology as it was mostly used to create code, document, and manage the graft project while I tested, architected, and mixed. I was keeping that to myself since maybe it'd actually be nice to have something to myself but after a handful of ignorant comments I'm open sourcing everything from it except the scripts themselves. You can hand the md back to claude or any competent agent and graft in the same way or have them explain it until you understand. No more dumbassery allowed now.
This is an evolving project subject to future fixes in H3 training. Training is clearly problematic, and my branch will rely on Sulphur H3 tunes. But I didn't want to wait for all that so I took the data from older models, grafting it in where the model needs it, grafting it to attn layers at a low level that doesn't disturb H3's visual or audio output quality. That's essentialy it.
minimax-h3-community-license-agreement). Because this release now carries transferred character from LTX 2.3, Wan 2.2, and Krea 2, the community licenses for those source models apply as well to the portions of character that came from each.