| File | minimax_h3_fl2va_bf16.safetensors |
| Model | MiniMax H3 (33B) (video model, 33.1B) |
| Precision | 16-bit |
| Size | 66.3 GB |
| VRAM to fit | about 72.1 GB |
| Folder | ComfyUI/models/diffusion_models/ |
| Loader node | Load Diffusion Model (UNETLoader) |
| Source | Comfy-Org/MiniMax-H3 on Hugging Face |
Download minimax_h3_fl2va_bf16.safetensors (66.3 GB)
Where to put it
Put the file in ComfyUI/models/diffusion_models/ (portable version: ComfyUI_windows_portable/ComfyUI/models/diffusion_models/). Older guides say models/unet: that folder still works. Press R in ComfyUI or restart it, then pick the file in Load Diffusion Model (UNETLoader).
If a workflow says “Value not in list” for that node, the workflow expects a file with another name. Click the node and choose this one.
Which graphics cards it fits
| VRAM | This file | Best file for that card (RTX 40/50) | |
|---|---|---|---|
| 6 GB | Not practical | Short by 66.1 GB | MiniMax-H3-FL2VA-Q3_K_M.gguf (15.6 GB) |
| 8 GB | Not practical | Short by 64.1 GB | MiniMax-H3-FL2VA-Q3_K_M.gguf (15.6 GB) |
| 10 GB | Not practical | Short by 62.1 GB | MiniMax-H3-FL2VA-Q4_K_M.gguf (19.9 GB) |
| 12 GB | Not practical | Short by 60.1 GB | MiniMax-H3-FL2VA-Q4_K_M.gguf (19.9 GB) |
| 16 GB | Not practical | Short by 56.1 GB | MiniMax-H3-FL2VA-Q4_K_M.gguf (19.9 GB) |
| 20 GB | Not practical | Short by 52.1 GB | MiniMax-H3-FL2VA-Q3_K_M.gguf (15.6 GB) |
| 24 GB | Not practical | Short by 48.1 GB | MiniMax-H3-FL2VA-Q3_K_M.gguf (15.6 GB) |
| 32 GB | Not practical | Short by 40.1 GB | MiniMax-H3-FL2VA-Q5_K_M.gguf (23.9 GB) |
| 48 GB | Offload only | Short by 24.1 GB: runs with part of it in system RAM, slower | minimax_h3_fl2va_int8_convrot.safetensors (34.0 GB) |
Need = file × 1 + 5 GB for the picture and 0.8 GB for Windows and the browser — the same rule as the rest of the site.
What else you need
This file alone does not make a picture. With it you load:
| Role | File | Size | Folder |
|---|---|---|---|
| Model | minimax_h3_fl2va_bf16.safetensors | 66.3 GB | models/diffusion_models |
| Text encoder | qwen3vl_32b_minimax_h3_int8_convrot.safetensors | 27.1 GB | models/text_encoders |
| VAE | minimax_h3_video_vae_fp16.safetensors | 5.2 GB | models/vae |
| Also needed | minimax_h3_audio_vae_fp32.safetensors | 0.6 GB | models/vae |
99.2 GB in total. Smaller text-encoder versions and other options are on the MiniMax H3 page.
Other files of MiniMax H3
| File | Precision | Size | VRAM to fit |
|---|---|---|---|
| minimax_h3_fl2va_bf16.safetensors | 16-bit | 66.3 GB | 72.1 GB |
| MiniMax-H3-FL2VA-Q8_0.gguf | Q8_0 | 36.0 GB | 41.8 GB |
| minimax_h3_fl2va_int8_convrot.safetensors | INT8 | 34.0 GB | 39.8 GB |
| MiniMax-H3-FL2VA-Q6_K.gguf | Q6_K | 28.2 GB | 34.0 GB |
| MiniMax-H3-FL2VA-Q5_K_M.gguf | Q5_K_M | 23.9 GB | 29.7 GB |
| MiniMax-H3-FL2VA-Q4_K_M.gguf | Q4_K_M | 19.9 GB | 25.7 GB |
| MiniMax-H3-FL2VA-Q3_K_M.gguf | Q3_K_M | 15.6 GB | 21.4 GB |
Same model, different compression. 8-bit files (FP8, INT8, Q8_0) look almost the same as 16-bit; below Q5 the loss starts to show.
Want the answer for your exact card? Check your GPU. Every file of every model: ComfyUI model files.
Size and source read from Hugging Face (2026-10-01).