T2V-A14B MoE video generation
According to the model card, the T2V-A14B model generates 5-second videos at 480P and 720P using a two-expert MoE design with about 14B active parameters per step.
Open Source Model Profile · Wan-AI
Wan2.2-T2V-A14B-Diffusers
Wan2.2-T2V-A14B-Diffusers is a 14.29B-parameter Wan-AI text-to-video MoE model. According to the model card, it generates 5-second videos at 480P and 720P.
- Publisher
- Wan-AI
- Task
- text-to-video
- Model type
- Unknown
- License
- apache-2.0
- Library
- diffusers
- Publication status
- Accepted · not indexed
Model overview
Wan2.2-T2V-A14B-Diffusers is published by Wan-AI as a text-to-video model in the Wan2.2 family. Safetensors metadata reports 14,288,491,584 parameters, and the record lists the diffusers library with an Apache-2.0 license. According to the model card, this repository holds the T2V-A14B MoE model for 5-second video generation at 480P and 720P.
Recorded capabilities
Publisher-described motion and data scale
According to the model card, Wan2.2 adds MoE capacity, curated aesthetic data, and training-data expansion of 65.6% more images and 83.2% more videos over Wan2.1.
Diffusers integration
According to the model card, the model loads as WanPipeline in bfloat16 with documented height, width, frame-count, and guidance settings, requiring diffusers features from the main branch.
Documented single- and multi-GPU paths
According to the model card, single-GPU 1280x720 inference needs at least 80GB VRAM, with memory-reduction flags and multi-GPU FSDP plus DeepSpeed Ulysses commands documented.
Use cases in the source record
- Text-to-video generation at 480P and 720P from prompts, following the model card's Diffusers workflow with documented frame counts and guidance scales.
- Single- or multi-GPU local inference experiments using the documented generate.py commands, subject to the publisher's 80GB VRAM and distributed-setup guidance.
Limitations and unknowns
- No evaluation results were extracted from this record beyond publisher-described Wan-Bench 2.0 and validation-loss comparisons.
- According to the model card, use requires installing diffusers from source because the needed features are only in the main branch.
- Provider state is historical snapshot data, not independently refreshed current availability.
Source and provenance
Source: Wan-AI/Wan2.2-T2V-A14B-Diffusers
Captured: Unknown. Processed: 2026-09-07T19:35:53.121539+00:00.
Wan2.2 💜 Wan ｜ 🖥️ GitHub | 🤗 Hugging Face | 🤖 ModelScope | 📑 Technical Report | 📑 Blog | 💬 WeChat Group | 📖 Discord Wan: Open and Advanced Large-Scale Video Generative Models We are excited to introduce Wan2.2 , a major upgrade to our foundational video models. With Wan2.2 , we have focused on incorporating the following innovations: 👍 Effective MoE Architecture : Wan2.2 introduces a Mixture-of-Experts (MoE) architecture into video diffusion models. By separating the denoising process cross timesteps with specialized powerful expert models, this enlarges the overall model capacity while maintaining the same computational co…
F001F002F003F004F005F006F008F009F011F013F017F018F019F020F024F025F026F027F030