MiniMax H3 Pruned VRAM requirements
An official, smaller cut of MiniMax H3 published by Comfy-Org, made for cards that cannot hold the full 33B model.
TypeVideo
Parameterspruned 33B
Fits entirely from15 GB
8-bit or better from27 GB
01The files, and how much VRAM each needs
| File | Size | Needed | Min. VRAM | Quality | Source |
|---|---|---|---|---|---|
| 16-bit | 40.2 GB | 46.0 GB | 47 GB | the original weights | Comfy-Org/MiniMax-H3 → |
| Q8_0 | 21.6 GB | 27.4 GB | 28 GB | practically identical to the original | Abiray/MiniMax-H3-Pruned-GGUF → |
| FP8 | 21.0 GB | 26.8 GB | 27 GB | practically identical to the original | Comfy-Org/MiniMax-H3 → |
| INT8 | 21.0 GB | 26.8 GB | 27 GB | practically identical to the original | Comfy-Org/MiniMax-H3 → |
| Q6_K | 16.7 GB | 22.5 GB | 23 GB | very close to the original | Abiray/MiniMax-H3-Pruned-GGUF → |
| Q5_K_M | 14.1 GB | 19.9 GB | 20 GB | close; small differences in fine detail | Abiray/MiniMax-H3-Pruned-GGUF → |
| Q4_K_M | 11.6 GB | 17.4 GB | 18 GB | good; some loss in fine detail and text | Abiray/MiniMax-H3-Pruned-GGUF → |
| Q3_K_M | 8.9 GB | 14.7 GB | 15 GB | noticeable loss of detail | Abiray/MiniMax-H3-Pruned-GGUF → |
“Needed” = file + 5 GB working memory + 0.8 GB system reserve.
02By amount of VRAM
03Best GPU for MiniMax H3 Pruned
The cheapest cards (by launch price) that run it well, and every card sorted by memory: best GPU for MiniMax H3 Pruned → Planning bigger images or longer clips? Open the calculator →
04By graphics card
| GPU | VRAM | Verdict | Best file | Needed |
|---|---|---|---|---|
| Desktop graphics cards | ||||
| RTX 2060 6 GB | 6 GB | Offload only | Q4_K_M | 17.4 GB |
| RTX 3050 6 GB | 6 GB | Offload only | Q4_K_M | 17.4 GB |
| RTX 2070 Super 8 GB | 8 GB | Offload only | Q4_K_M | 17.4 GB |
| RTX 2080 Super 8 GB | 8 GB | Offload only | Q4_K_M | 17.4 GB |
| RTX 3050 8 GB | 8 GB | Offload only | Q4_K_M | 17.4 GB |
| RTX 3060 8 GB | 8 GB | Offload only | Q4_K_M | 17.4 GB |
| RTX 3060 Ti 8 GB | 8 GB | Offload only | Q4_K_M | 17.4 GB |
| RTX 3070 8 GB | 8 GB | Offload only | Q4_K_M | 17.4 GB |
| RTX 3070 Ti 8 GB | 8 GB | Offload only | Q4_K_M | 17.4 GB |
| RTX 4060 8 GB | 8 GB | Offload only | Q4_K_M | 17.4 GB |
| RTX 4060 Ti 8 GB | 8 GB | Offload only | Q4_K_M | 17.4 GB |
| RTX 5050 8 GB | 8 GB | Offload only | Q4_K_M | 17.4 GB |
| RTX 5060 8 GB | 8 GB | Offload only | Q4_K_M | 17.4 GB |
| RTX 5060 Ti 8 GB | 8 GB | Offload only | Q4_K_M | 17.4 GB |
| RX 7600 8 GB | 8 GB | Offload only | Q4_K_M | 17.4 GB |
| RX 9050 8 GB | 8 GB | Offload only | Q4_K_M | 17.4 GB |
| RX 9060 XT 8 GB | 8 GB | Offload only | Q4_K_M | 17.4 GB |
| Arc B570 10 GB | 10 GB | Offload only | Q4_K_M | 17.4 GB |
| RTX 3080 10 GB | 10 GB | Offload only | Q4_K_M | 17.4 GB |
| RTX 2080 Ti 11 GB | 11 GB | Offload only | Q4_K_M | 17.4 GB |
| Arc B580 12 GB | 12 GB | Offload only | Q4_K_M | 17.4 GB |
| RTX 2060 12 GB | 12 GB | Offload only | Q4_K_M | 17.4 GB |
| RTX 3060 12 GB | 12 GB | Offload only | Q4_K_M | 17.4 GB |
| RTX 3080 12 GB | 12 GB | Offload only | Q4_K_M | 17.4 GB |
| RTX 3080 Ti 12 GB | 12 GB | Offload only | Q4_K_M | 17.4 GB |
| RTX 4070 12 GB | 12 GB | Offload only | Q4_K_M | 17.4 GB |
| RTX 4070 Super 12 GB | 12 GB | Offload only | Q4_K_M | 17.4 GB |
| RTX 4070 Ti 12 GB | 12 GB | Offload only | Q4_K_M | 17.4 GB |
| RTX 5070 12 GB | 12 GB | Offload only | Q4_K_M | 17.4 GB |
| RX 7700 XT 12 GB | 12 GB | Offload only | Q4_K_M | 17.4 GB |
| RX 9070 GRE 12 GB | 12 GB | Offload only | Q4_K_M | 17.4 GB |
| Arc A770 16 GB | 16 GB | Tight | Q3_K_M | 14.7 GB |
| RTX 4060 Ti 16 GB | 16 GB | Tight | Q3_K_M | 14.7 GB |
| RTX 4070 Ti Super 16 GB | 16 GB | Tight | Q3_K_M | 14.7 GB |
| RTX 4080 16 GB | 16 GB | Tight | Q3_K_M | 14.7 GB |
| RTX 4080 Super 16 GB | 16 GB | Tight | Q3_K_M | 14.7 GB |
| RTX 5060 Ti 16 GB | 16 GB | Tight | Q3_K_M | 14.7 GB |
| RTX 5070 Ti 16 GB | 16 GB | Tight | Q3_K_M | 14.7 GB |
| RTX 5080 16 GB | 16 GB | Tight | Q3_K_M | 14.7 GB |
| RX 7600 XT 16 GB | 16 GB | Tight | Q3_K_M | 14.7 GB |
| RX 7800 XT 16 GB | 16 GB | Tight | Q3_K_M | 14.7 GB |
| RX 7900 GRE 16 GB | 16 GB | Tight | Q3_K_M | 14.7 GB |
| RX 9060 XT 16 GB | 16 GB | Tight | Q3_K_M | 14.7 GB |
| RX 9070 16 GB | 16 GB | Tight | Q3_K_M | 14.7 GB |
| RX 9070 XT 16 GB | 16 GB | Tight | Q3_K_M | 14.7 GB |
| RX 7900 XT 20 GB | 20 GB | Runs | Q5_K_M | 19.9 GB |
| Arc Pro B60 24 GB | 24 GB | Runs | Q6_K | 22.5 GB |
| RTX 3090 24 GB | 24 GB | Runs | Q6_K | 22.5 GB |
| RTX 3090 Ti 24 GB | 24 GB | Runs | Q6_K | 22.5 GB |
| RTX 4090 24 GB | 24 GB | Runs | Q6_K | 22.5 GB |
| RX 7900 XTX 24 GB | 24 GB | Runs | Q6_K | 22.5 GB |
| Arc Pro B70 32 GB | 32 GB | Runs well | Q8_0 | 27.4 GB |
| RTX 5090 32 GB | 32 GB | Runs well | FP8 | 26.8 GB |
| Laptop GPUs | ||||
| RTX 3050 Laptop 4 GB | 4 GB | Not practical | Q3_K_M | 14.7 GB |
| RTX 3050 Ti Laptop 4 GB | 4 GB | Not practical | Q3_K_M | 14.7 GB |
| RTX 2060 Laptop 6 GB | 6 GB | Offload only | Q4_K_M | 17.4 GB |
| RTX 3050 Laptop 6 GB | 6 GB | Offload only | Q4_K_M | 17.4 GB |
| RTX 3060 Laptop 6 GB | 6 GB | Offload only | Q4_K_M | 17.4 GB |
| RTX 4050 Laptop 6 GB | 6 GB | Offload only | Q4_K_M | 17.4 GB |
| RTX 2070 Laptop 8 GB | 8 GB | Offload only | Q4_K_M | 17.4 GB |
| RTX 2070 Super Laptop 8 GB | 8 GB | Offload only | Q4_K_M | 17.4 GB |
| RTX 2080 Laptop 8 GB | 8 GB | Offload only | Q4_K_M | 17.4 GB |
| RTX 2080 Super Laptop 8 GB | 8 GB | Offload only | Q4_K_M | 17.4 GB |
| RTX 3070 Laptop 8 GB | 8 GB | Offload only | Q4_K_M | 17.4 GB |
| RTX 3070 Ti Laptop 8 GB | 8 GB | Offload only | Q4_K_M | 17.4 GB |
| RTX 3080 Laptop 8 GB | 8 GB | Offload only | Q4_K_M | 17.4 GB |
| RTX 4060 Laptop 8 GB | 8 GB | Offload only | Q4_K_M | 17.4 GB |
| RTX 4070 Laptop 8 GB | 8 GB | Offload only | Q4_K_M | 17.4 GB |
| RTX 5050 Laptop 8 GB | 8 GB | Offload only | Q4_K_M | 17.4 GB |
| RTX 5060 Laptop 8 GB | 8 GB | Offload only | Q4_K_M | 17.4 GB |
| RTX 5070 Laptop 8 GB | 8 GB | Offload only | Q4_K_M | 17.4 GB |
| RX 7600M 8 GB | 8 GB | Offload only | Q4_K_M | 17.4 GB |
| RX 7600M XT 8 GB | 8 GB | Offload only | Q4_K_M | 17.4 GB |
| RX 7600S 8 GB | 8 GB | Offload only | Q4_K_M | 17.4 GB |
| RX 7700S 8 GB | 8 GB | Offload only | Q4_K_M | 17.4 GB |
| RTX 4080 Laptop 12 GB | 12 GB | Offload only | Q4_K_M | 17.4 GB |
| RTX 5070 Laptop 12 GB | 12 GB | Offload only | Q4_K_M | 17.4 GB |
| RTX 5070 Ti Laptop 12 GB | 12 GB | Offload only | Q4_K_M | 17.4 GB |
| RX 7800M 12 GB | 12 GB | Offload only | Q4_K_M | 17.4 GB |
| RTX 3080 Laptop 16 GB | 16 GB | Tight | Q3_K_M | 14.7 GB |
| RTX 3080 Ti Laptop 16 GB | 16 GB | Tight | Q3_K_M | 14.7 GB |
| RTX 4090 Laptop 16 GB | 16 GB | Tight | Q3_K_M | 14.7 GB |
| RTX 5080 Laptop 16 GB | 16 GB | Tight | Q3_K_M | 14.7 GB |
| RX 7900M 16 GB | 16 GB | Tight | Q3_K_M | 14.7 GB |
| RTX 5090 Laptop 24 GB | 24 GB | Runs | Q6_K | 22.5 GB |
| Unified memory | ||||
| Radeon 8060S (Strix Halo) 96 GB | 96 GB | Runs well | 16-bit | 46.0 GB |
| Radeon 8065S (Gorgon Halo) 160 GB | 160 GB | Runs well | 16-bit | 46.0 GB |
| Apple Silicon Macs (by memory) | ||||
| Mac 16 GB | 12.7 GB | Offload only | Q4_K_M | 17.4 GB |
| Mac 18 GB | 14.4 GB | Offload only | Q3_K_M | 14.7 GB |
| Mac 24 GB | 19.6 GB | Runs | Q4_K_M | 17.4 GB |
| Mac 32 GB | 26.8 GB | Runs | Q6_K | 22.5 GB |
| Mac 36 GB | 30.2 GB | Runs well | Q8_0 | 27.4 GB |
| Mac 48 GB | 40.2 GB | Runs well | Q8_0 | 27.4 GB |
| Mac 64 GB | 55.7 GB | Runs well | 16-bit | 46.0 GB |
| Mac 96 GB | 85 GB | Runs well | 16-bit | 46.0 GB |
| Mac 128 GB | 115.4 GB | Runs well | 16-bit | 46.0 GB |
| Mac 192 GB | 175.4 GB | Runs well | 16-bit | 46.0 GB |
| Mac 256 GB | 236.9 GB | Runs well | 16-bit | 46.0 GB |
| Mac 512 GB | 498.1 GB | Runs well | 16-bit | 46.0 GB |
05Text encoder, VAE and other files
| File | Folder | Size | When | |
|---|---|---|---|---|
| Qwen3-VL 32B NVFP4 (AWQ) qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors | models/text_encoders | 15.7 GB | text encoder · smaller, recommended | Download → |
| Qwen3-VL 32B INT8 (convrot) qwen3vl_32b_minimax_h3_int8_convrot.safetensors | models/text_encoders | 27.1 GB | text encoder · alternative | Download → |
| Qwen3-VL 32B BF16 qwen3vl_32b_minimax_h3_bf16.safetensors | models/text_encoders | 51.5 GB | text encoder · alternative | Download → |
| MiniMax H3 video VAE FP16 minimax_h3_video_vae_fp16.safetensors | models/vae | 5.2 GB | required | Download → |
| MiniMax H3 audio VAE FP32 minimax_h3_audio_vae_fp32.safetensors | models/vae | 0.6 GB | required | Download → |
The files the official ComfyUI workflows load next to the model. Sizes read from Hugging Face (2026-10-01).
At 15.7 GB even the smallest Qwen3-VL 32B file is too big for 16 GB, so it runs from system RAM. That slows down reading the prompt, not the picture. Sizes: 16-bit 51.5 GB, INT8 27.1 GB, NVFP4 15.7 GB. Why the encoder does not count against VRAM →
06Where the files go in ComfyUI
| File | Folder | Loader node |
|---|---|---|
| Diffusion model (.safetensors: 16-bit, FP8, INT8) | ComfyUI/models/diffusion_models | Load Diffusion Model |
| GGUF file (.gguf) | ComfyUI/models/unet | Unet Loader (GGUF) - from the ComfyUI-GGUF node pack |
| Text encoder | ComfyUI/models/text_encoders | Load CLIP / DualCLIPLoader (or the GGUF versions) |
| VAE | ComfyUI/models/vae | Load VAE |
07Measured and reported results
| Label | GPU | Setup | Result | Peak VRAM | Date | Source |
|---|---|---|---|---|---|---|
| reported | RTX 4060 8 GB | FL2VA pruned 20B INT8 ConvRot + Q2_K GGUF text encoder · 832x480 · 20 steps “Denoising time: About 5.5 minutes” Desktop RTX 4060 8GB, 96 GB RAM, WSL; 5.5 min denoise only; "Total first-run time: Just under 8 minutes"; a later 15-second run took about 15 minutes | 330 s / clip (107 frames) | — | 2026-08 | mountainmeadowsystems.com → |
| reported | RTX 4070 Laptop 8 GB | pruned int8 convrot (FL2VA/Ref2VA) · 832x480 “On a laptop with an RTX 4070 (8GB VRAM) and 32GB of RAM, we generated one 832×480, 15-second clip in 45 minutes.” 15-second clip, 24 fps, with audio; ComfyUI I2V/R2V workflows (which one not stated); 45 min = 2700 s; steps not stated | 2700 s / clip | — | 2026-08-04 | metallab.ai → |
| reported | RTX 4090 24 GB | minimax_h3_fl2va_pruned_int8_convrot + qwen3vl_32b nvfp4_awq · 832x480 · 8 steps “5 s clip (832×480, 8 steps): ~7 min” ~7 min approx = 420 s for 5 s clip; VRAM ~6.6 GB during sampling, ~22.5 GB spike at model load; 15 s clip ~25-30 min; no date shown | 420 s / clip | 22.5 GB | — | github.com → |
| reported | RTX 5090 32 GB | minimax_h3_fl2va_pruned_nvfp4 · 864x480 · 10 steps “175 s for a 864×480 ten-second clip, 26,914 MiB peak VRAM.” ComfyUI 0.30.1, 500 W power cap; 26,914 MiB = 26.28 GiB; INT8-ConvRot = 185 s / 28,581 MiB | 175 s / clip (243 frames) | 26.28 GB | 2026-08-04 | ai-muninn.com → |
Other people's numbers are copied as published, with a link. Send yours.
08Training a LoRA for MiniMax H3 Pruned
No trainer documentation with a VRAM figure for this model was found yet. What the trainers say for other models →
09Every file tracked for MiniMax H3 Pruned
| File | Type | Size | Repo |
|---|---|---|---|
| MiniMax-H3-FL2VA-Pruned-Q8_0.gguf (fl2va-pruned) | Q8_0 | 21.58 GB | Abiray/MiniMax-H3-Pruned-GGUF → |
| MiniMax-H3-Ref2VA-Pruned-Q8_0.gguf (ref2va-pruned) | Q8_0 | 21.58 GB | Abiray/MiniMax-H3-Pruned-GGUF → |
| MiniMax-H3-FL2VA-Pruned-Q6_K.gguf (fl2va-pruned) | Q6_K | 16.73 GB | Abiray/MiniMax-H3-Pruned-GGUF → |
| MiniMax-H3-Ref2VA-Pruned-Q6_K.gguf (ref2va-pruned) | Q6_K | 16.73 GB | Abiray/MiniMax-H3-Pruned-GGUF → |
| MiniMax-H3-FL2VA-Pruned-Q5_K_M.gguf (fl2va-pruned) | Q5_K_M | 14.07 GB | Abiray/MiniMax-H3-Pruned-GGUF → |
| MiniMax-H3-FL2VA-Pruned-Q5_K_S.gguf (fl2va-pruned) | Q5_K_S | 14.07 GB | Abiray/MiniMax-H3-Pruned-GGUF → |
| MiniMax-H3-Ref2VA-Pruned-Q5_K_M.gguf (ref2va-pruned) | Q5_K_M | 14.07 GB | Abiray/MiniMax-H3-Pruned-GGUF → |
| MiniMax-H3-Ref2VA-Pruned-Q5_K_S.gguf (ref2va-pruned) | Q5_K_S | 14.07 GB | Abiray/MiniMax-H3-Pruned-GGUF → |
| MiniMax-H3-FL2VA-Pruned-Q4_K_M.gguf (fl2va-pruned) | Q4_K_M | 11.56 GB | Abiray/MiniMax-H3-Pruned-GGUF → |
| MiniMax-H3-FL2VA-Pruned-Q4_K_S.gguf (fl2va-pruned) | Q4_K_S | 11.56 GB | Abiray/MiniMax-H3-Pruned-GGUF → |
| MiniMax-H3-Ref2VA-Pruned-Q4_K_M.gguf (ref2va-pruned) | Q4_K_M | 11.56 GB | Abiray/MiniMax-H3-Pruned-GGUF → |
| MiniMax-H3-Ref2VA-Pruned-Q4_K_S.gguf (ref2va-pruned) | Q4_K_S | 11.56 GB | Abiray/MiniMax-H3-Pruned-GGUF → |
| MiniMax-H3-FL2VA-Pruned-Q3_K_M.gguf (fl2va-pruned) | Q3_K_M | 8.90 GB | Abiray/MiniMax-H3-Pruned-GGUF → |
| MiniMax-H3-Ref2VA-Pruned-Q3_K_M.gguf (ref2va-pruned) | Q3_K_M | 8.90 GB | Abiray/MiniMax-H3-Pruned-GGUF → |
| minimax_h3_fl2va_pruned_bf16.safetensors (fl2va-pruned) | BF16 | 40.23 GB | Comfy-Org/MiniMax-H3 → |
| minimax_h3_ref2va_pruned_bf16.safetensors (ref2va-pruned) | BF16 | 40.23 GB | Comfy-Org/MiniMax-H3 → |
| minimax_h3_fl2va_pruned_int8_convrot.safetensors (fl2va-pruned) | INT8 | 20.97 GB | Comfy-Org/MiniMax-H3 → |
| minimax_h3_ref2va_pruned_int8_convrot.safetensors (ref2va-pruned) | INT8 | 20.97 GB | Comfy-Org/MiniMax-H3 → |
| minimax_h3_fl2va_pruned_fp8_scaled.safetensors (fl2va-pruned) | FP8 | 20.96 GB | Comfy-Org/MiniMax-H3 → |
| minimax_h3_ref2va_pruned_fp8_scaled.safetensors (ref2va-pruned) | FP8 | 20.96 GB | Comfy-Org/MiniMax-H3 → |
33B (HF API 33.1B) omni transformer generating video WITH native stereo audio, 4-15 s, 768p native (2K via upscale pass). Two denoisers: fl2va (text/first/last-frame-to-video+audio) and ref2va (up to 9 images/3 videos/3 audio refs); load one. 'Pruned' variants are smaller official Comfy-Org cuts (40.2 GB bf16 vs 66.3 GB). VAEs: minimax_h3_video_vae_fp16 5207808496 + minimax_h3_audio_vae_fp32 605254808. Native ComfyUI support (day-0, Aug 2026 blog); Comfy blog says smallest variants total 42.5 GB (from 123.6 GB). unsloth/MiniMax-H3-GGUF (~1.18M downloads, pruned only, quants named Q3_K/Q4_K/UD-*) is documented for stable-diffusion.cpp, not ComfyUI - not listed.