How much VRAM does Wan 2.2 TI2V 5B need?

At BF16 with everything resident, Wan 2.2 TI2V 5B peaks at about 24.61 GB. Offloading components one at a time brings that to 14.02 GB, and quantising the weights to 4-bit brings it to 10.18 GB.

What is the cheapest GPU that runs Wan 2.2 TI2V 5B?

The GeForce GTX 1660 SUPER at around $110, running it at GGUF Q8_0 with sequential offload, taking about 20 h 22 min per generation.

Wan · Diffusion transformer

Wan 2.2 TI2V 5B

Wan 2.2 TI2V 5B is a pipeline of 3 networks totalling 11.386 billion parameters, of which 5 billion do the actual generating. At its native 1280x704 and 121 frames the denoiser works on 27,280 latent tokens at once, and attention over that sequence is what sets the render time.

The one video model most people can actually run: five billion parameters and a 720p output.

Whole pipeline11.386B

Denoiser5B

Download31.8 GB

Default steps40

Native length121 frames

Frame rate24 fps

What the pipeline is made of

This is the part that separates a media model from a language model: these are separate networks, and whether they sit in memory together is your choice, not the model's.

Component	Parameters	Published as	On disk	What it does
transformer	5B	F32	18.63 GB	Does the generating. Runs once per step, so it dominates both memory and time.
text_encoder	5.681B	BF16	10.58 GB	Turns your prompt into conditioning. Runs once, then can leave the GPU entirely.
vae	0.705B	F32	2.63 GB	Converts between pixels and the compressed latent space. Small, but its decode pass is a memory spike.

These figures were measured from the weight file headers.

Memory by precision

Everything resident, at the model's native output size, on a card large enough that precision is the only constraint.

Precision	Peak VRAM	Quality	What it costs you
BF16	24.61 GB	100%	How the weights are published. No loss, and the largest footprint.
FP8	14.66 GB	97%	Halves the denoiser with a small, usually invisible cost. Needs Ada or newer.
GGUF Q8_0	15.28 GB	98%	Works on any card, unlike FP8. Slightly slower than native precision.
GGUF Q5_K_M	11.78 GB	95%	A middle step when Q8 will not fit.
GGUF Q4_K_M	10.72 GB	91%	The usual way a 12B image model gets onto an 8 GB card. Detail softens.
NF4	10.18 GB	89%	Aggressive 4-bit. Fast to load, noticeably looser on fine detail.

Memory by arrangement

The other lever: the same weights at the same precision, moved around differently. On this model quantising is the stronger lever: the denoiser is 44% of the pipeline and stays resident whatever you rearrange.

Arrangement	Peak VRAM	Time	How it works
Everything resident	24.61 GB	6 min 20 s	All components stay on the GPU. Fastest, and needs the most memory.
Text encoder on CPU	14.02 GB	6 min 38 s	The prompt is encoded once on the processor, so the encoder never touches the GPU at all. Standard practice on video models, whose encoders are often larger than their denoisers.
Component offload	13.98 GB	7 min 5 s	One component on the GPU at a time. The text encoder runs, then makes way for the denoiser. Costs a few seconds per generation.
Sequential offload	5.78 GB	37 min 30 s	Weights stream layer by layer from system RAM. Runs almost anything on almost anything, and is many times slower.

What a bigger output costs

Both memory and time climb faster than the frame count, because attention is quadratic in the length of the latent sequence.

Output	Latent tokens	Peak VRAM	Time
480p, 3 seconds	5,070	22.98 GB	28 s
480p, 5 seconds	8,190	22.98 GB	55 s
720p, 3 seconds	11,960	23.2 GB	1 min 37 s
720p, 5 seconds	19,320	23.88 GB	3 min 31 s
720p, 10 seconds	37,720	25.56 GB	11 min 13 s

Measured at BF16 with everything resident on an RTX 4090, so the columns compare with each other rather than with your machine.

Which hardware runs Wan 2.2 TI2V 5B

117 of 118 consumer devices run it in some arrangement. Each row shows the fastest arrangement that fits on that device.

Device	Memory	Verdict	Arrangement	Peak	Time
RTX PRO 6000 Blackwell NVIDIA	96 GB	Workable	BF16 everything resident	24.61 GB	4 min 11 s
GeForce RTX 5090 NVIDIA	32 GB	Slow	BF16 everything resident	24.61 GB	5 min 0 s
RTX 6000 Ada Generation NVIDIA	48 GB	Slow	BF16 everything resident	24.61 GB	5 min 43 s
GeForce RTX 4090 NVIDIA	24 GB	Slow	GGUF Q8_0 everything resident	15.28 GB	6 min 20 s
GeForce RTX 5090 Laptop NVIDIA	24 GB	Slow	GGUF Q8_0 everything resident	15.28 GB	8 min 0 s
NVIDIA DGX Spark 128GB NVIDIA	128 GB	Slow	BF16 everything resident	24.61 GB	8 min 19 s
GeForce RTX 5080 NVIDIA	16 GB	Slow	FP8 everything resident	14.66 GB	9 min 14 s
GeForce RTX 4080 SUPER NVIDIA	16 GB	Slow	FP8 everything resident	14.66 GB	9 min 59 s
GeForce RTX 4090 Laptop NVIDIA	16 GB	Slow	FP8 everything resident	14.66 GB	10 min 23 s
GeForce RTX 4080 NVIDIA	16 GB	Slow	FP8 everything resident	14.66 GB	10 min 39 s
GeForce RTX 5080 Laptop NVIDIA	16 GB	Slow	FP8 everything resident	14.66 GB	10 min 55 s
GeForce RTX 5070 Ti NVIDIA	16 GB	Slow	FP8 everything resident	14.66 GB	11 min 43 s
GeForce RTX 4070 Ti SUPER NVIDIA	16 GB	Slow	FP8 everything resident	14.66 GB	11 min 47 s
GeForce RTX 4070 Ti NVIDIA	12 GB	Slow	GGUF Q4_K_M everything resident	10.72 GB	12 min 57 s
GeForce RTX 3090 Ti NVIDIA	24 GB	Slow	GGUF Q8_0 everything resident	15.28 GB	12 min 57 s
RTX A6000 NVIDIA	48 GB	Slow	BF16 everything resident	24.61 GB	13 min 22 s
GeForce RTX 4080 Laptop NVIDIA	12 GB	Slow	GGUF Q4_K_M everything resident	10.72 GB	13 min 60 s
GeForce RTX 3090 NVIDIA	24 GB	Slow	GGUF Q8_0 everything resident	15.28 GB	14 min 35 s
GeForce RTX 4070 SUPER NVIDIA	12 GB	Slow	GGUF Q4_K_M everything resident	10.72 GB	14 min 41 s
GeForce RTX 3080 Ti NVIDIA	12 GB	Slow	GGUF Q4_K_M everything resident	10.72 GB	15 min 13 s
GeForce RTX 5070 NVIDIA	12 GB	Slow	GGUF Q4_K_M everything resident	10.72 GB	16 min 49 s
GeForce RTX 3080 12GB NVIDIA	12 GB	Slow	GGUF Q4_K_M everything resident	10.72 GB	17 min 23 s
GeForce RTX 4070 NVIDIA	12 GB	Slow	GGUF Q4_K_M everything resident	10.72 GB	17 min 50 s
GeForce RTX 3080 10GB NVIDIA	10 GB	Slow	GGUF Q5_K_M text encoder on cpu	8.02 GB	18 min 15 s
RTX A5000 NVIDIA	24 GB	Slow	GGUF Q8_0 everything resident	15.28 GB	18 min 38 s
GeForce RTX 2080 Ti NVIDIA	11 GB	Slow	NF4 everything resident	10.18 GB	19 min 9 s
GeForce RTX 5060 Ti 16GB NVIDIA	16 GB	Barely usable	FP8 everything resident	14.66 GB	21 min 32 s
GeForce RTX 4060 Ti 16GB NVIDIA	16 GB	Barely usable	FP8 everything resident	14.66 GB	23 min 28 s
GeForce RTX 5060 Ti 8GB NVIDIA	8 GB	Barely usable	GGUF Q5_K_M component offload	7.16 GB	24 min 6 s
Jetson AGX Orin 32GB NVIDIA	32 GB	Barely usable	GGUF Q8_0 everything resident	15.28 GB	24 min 18 s
Jetson AGX Orin 64GB NVIDIA	64 GB	Barely usable	BF16 everything resident	24.61 GB	24 min 18 s
Radeon RX 7900 XTX AMD	24 GB	Barely usable	GGUF Q8_0 everything resident	15.28 GB	24 min 29 s
Radeon PRO W7900 AMD	48 GB	Barely usable	BF16 everything resident	24.61 GB	24 min 41 s
GeForce RTX 4070 Laptop NVIDIA	8 GB	Barely usable	GGUF Q5_K_M component offload	7.16 GB	24 min 52 s
GeForce RTX 4060 Ti 8GB NVIDIA	8 GB	Barely usable	GGUF Q5_K_M component offload	7.16 GB	26 min 17 s
GeForce RTX 3070 Ti NVIDIA	8 GB	Barely usable	GGUF Q5_K_M component offload	7.16 GB	26 min 35 s
RTX A4000 NVIDIA	16 GB	Barely usable	FP8 everything resident	14.66 GB	26 min 49 s
GeForce RTX 3070 NVIDIA	8 GB	Barely usable	GGUF Q5_K_M component offload	7.16 GB	28 min 33 s
Radeon RX 7900 XT AMD	20 GB	Barely usable	GGUF Q8_0 everything resident	15.28 GB	29 min 14 s
GeForce RTX 5060 NVIDIA	8 GB	Barely usable	GGUF Q5_K_M component offload	7.16 GB	30 min 1 s
Radeon RX 9070 XT AMD	16 GB	Barely usable	FP8 everything resident	14.66 GB	31 min 2 s
Radeon RX 7900 GRE AMD	16 GB	Barely usable	FP8 everything resident	14.66 GB	32 min 43 s
GeForce RTX 4060 Laptop NVIDIA	8 GB	Barely usable	GGUF Q5_K_M component offload	7.16 GB	35 min 1 s
GeForce RTX 3060 Ti NVIDIA	8 GB	Barely usable	GGUF Q5_K_M component offload	7.16 GB	35 min 33 s
GeForce RTX 4060 NVIDIA	8 GB	Barely usable	GGUF Q5_K_M component offload	7.16 GB	37 min 52 s

Source: Wan-AI/Wan2.2-TI2V-5B-Diffusers . Downloaded 0.2 million times in the last month. See how we calculate, or browse every video model.