Wan 2.2 VRAM requirements
Short answer: Wan-AI's own model cards say the TI2V-5B model can run on a GPU with at least 24 GB of VRAM (e.g. RTX 4090) when you use its offload options, and the T2V-A14B command "can run on a GPU with at least 80GB VRAM".[1][2]
Wan 2.2 is Wan-AI's open video generation family, licensed Apache 2.0. Both repos were created on 2025-07-28.[1][2]
Official VRAM figures
| Model | Official VRAM figure | Notes | Source |
|---|---|---|---|
| Wan2.2-TI2V-5B (text and image to video, 720P 24fps) | At least 24 GB (e.g. RTX 4090) | With --offload_model True --convert_model_dtype --t5_cpu. With 80 GB+ you can drop those flags to run faster. |
[1] |
| Wan2.2-T2V-A14B (text to video) | At least 80 GB | Single-GPU command. Two ~14B experts (about 27B total, 14B active per step). | [2] |
Weights only: file sizes on Hugging Face
Exact sizes from the Diffusers repos (1 GB = 1,000,000,000 bytes; split files added together; checked November 4, 2026).[1][2]
| Component | TI2V-5B | T2V-A14B |
|---|---|---|
| Transformer | 20.00 GB | 57.15 GB |
| Second transformer (low-noise expert) | — | 57.15 GB |
| Text encoder | 11.36 GB | 11.36 GB |
| VAE | 2.82 GB | 0.51 GB |
| Total (weights only) | 34.18 GB | 126.18 GB |
The A14B model only runs one expert per denoising step, which is why its memory need is much lower than the 126 GB of files.[2]
Which one should you use?
- 24 GB card (RTX 3090, 4090): TI2V-5B, using the offload flags from the model card.[1] See best GPUs with 24 GB.
- Smaller card: neither model card gives an official figure below 24 GB, so we don't claim one.
- No big GPU: rent an 80 GB GPU by the hour for A14B.
Frequently asked questions
How much VRAM does Wan 2.2 need?
The official TI2V-5B card says at least 24 GB (e.g. RTX 4090) with offload options. The T2V-A14B card says at least 80 GB for its single-GPU command.
Can Wan 2.2 run on an RTX 4090?
Yes for TI2V-5B, according to Wan-AI's model card, using the offload, dtype conversion and T5-on-CPU options.
How big are the Wan 2.2 model files?
TI2V-5B's Diffusers files total 34.18 GB. T2V-A14B's total 126.18 GB, because it ships two ~14B expert transformers.
Every number on this page links to its source.
How we calculate + sources ⓘ
Sources
All sources checked on .
- Hugging Face, Wan-AI/Wan2.2-TI2V-5B-Diffusers (model card and file listing; checked November 4, 2026). https://huggingface.co/Wan-AI/Wan2.2-TI2V-5B-Diffusers
- Hugging Face, Wan-AI/Wan2.2-T2V-A14B-Diffusers (model card and file listing; checked November 4, 2026). https://huggingface.co/Wan-AI/Wan2.2-T2V-A14B-Diffusers