Model reference · open weights
LingBot-Video-MoE is an open-weight video model from FastVideo. LingBot-Video-MoE-30B-A3B-Diffusers (BF16) weighs 130 GB; the smallest configuration that runs it is H100 80 GB.
Summary of the FastVideo/LingBot-Video-MoE-30B-A3B-Diffusers model card, 2026-10-01
What it is
| Released by | FastVideo |
|---|---|
| Released | 2026-07-13 |
| VRAM | 130 GB for the weights |
What it runs on
| Card | Weights | Memory |
|---|---|---|
| RTX 3060 12 GB … L40S 48 GB | does not fit | |
| A100 80 GB | fits (encoders offloaded) | 78.2 GB |
| H100 80 GB | fits (encoders offloaded) | 78.1 GB |
| RTX PRO 6000 Blackwell 96 GB | fits (encoders offloaded) | 93.8 GB |
| DGX Spark (GB10) 128 GB unified | fits (encoders offloaded) | 107 GB |
| H200 141 GB | tight | 138 GB |
| B200 180 GB | fits | 176 GB |
From the model card
This repository contains the released LingBot-Video MoE 30B-A3B base model and its second-stage refiner in the component layout loaded natively by FastVideo.
Use a FastVideo revision containing LingBot-Video support:
from fastvideo import VideoGenerator
generator = VideoGenerator.from_pretrained(
"FastVideo/LingBot-Video-MoE-30B-A3B-Diffusers",
num_gpus=8,
sp_size=8,
use_fsdp_inference=True,
refine_enabled=True,
text_encoder_cpu_offload=True,
)
result = generator.generate({
"prompt": "A red fox runs through fresh snow at sunrise.",
"output": {"output_path": "outputs/lingbot-video", "save_video": True},
})
generator.shutdown()
The released preset generates a 121-frame 832×480 base video and optionally
refines it to 1920×1088. Set refine_enabled=False when loading the model to run
the base MoE transformer without the refiner.
The original model and this converted package are distributed under the Apache-2.0 license declared by the released checkpoint.
Quoted from the model card on Hugging Face. The full card is behind the Hugging Face link above.
Running it yourself
Rent a machine by the hour. How to run this model is on its model card.