Wan ยท Diffusion transformer
Wan 2.2 T2V A14B
Wan 2.2 T2V A14B is a pipeline of 4 networks totalling 34.384 billion parameters, of which 14.288 billion do the actual generating. At its native 1280x720 and 81 frames the denoiser works on 75,600 latent tokens at once, and attention over that sequence is what sets the render time.
Whole pipeline34.384B
Denoiser14.288B
Download117.5 GB
Default steps40
Native length81 frames
Frame rate16 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 | 14.288B | F32 | 53.23 GB | Does the generating. Runs once per step, so it dominates both memory and time. |
| transformer_2 | 14.288B | F32 | 53.23 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 | 0.127B | F32 | 0.47 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 | 76.48 GB | 100% | How the weights are published. No loss, and the largest footprint. |
| FP8 | 44.58 GB | 97% | Halves the denoiser with a small, usually invisible cost. Needs Ada or newer. |
| GGUF Q8_0 | 46.57 GB | 98% | Works on any card, unlike FP8. Slightly slower than native precision. |
| GGUF Q5_K_M | 35.36 GB | 95% | A middle step when Q8 will not fit. |
| GGUF Q4_K_M | 31.94 GB | 91% | The usual way a 12B image model gets onto an 8 GB card. Detail softens. |
| NF4 | 30.22 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 42% of the pipeline and stays resident whatever you rearrange.
| Arrangement | Peak VRAM | Time | How it works |
|---|---|---|---|
| Everything resident | 76.48 GB | 1 h 13 min | All components stay on the GPU. Fastest, and needs the most memory. |
| Text encoder on CPU | 65.9 GB | 1 h 17 min | 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 | 65.66 GB | 1 h 22 min | 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 | 20.09 GB | 7 h 23 min | 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 | 20,280 | 68.04 GB | 7 min 47 s |
| 480p, 5 seconds | 32,760 | 69.94 GB | 16 min 57 s |
| 720p, 3 seconds | 46,800 | 72.09 GB | 31 min 17 s |
| 720p, 5 seconds | 75,600 | 76.48 GB | 1 h 13 min |
| 720p, 10 seconds | 147,600 | 87.47 GB | 4 h 18 min |
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 T2V A14B
74 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 | Barely usable | BF16 everything resident | 76.48 GB | 48 min 31 s |
| GeForce RTX 5090 NVIDIA | 32 GB | Barely usable | NF4 everything resident | 30.22 GB | 58 min 13 s |
| RTX 6000 Ada Generation NVIDIA | 48 GB | Barely usable | GGUF Q8_0 everything resident | 46.57 GB | 1 h 6 min |
| NVIDIA DGX Spark 128GB NVIDIA | 128 GB | Barely usable | BF16 everything resident | 76.48 GB | 1 h 37 min |
| RTX A6000 NVIDIA | 48 GB | Barely usable | GGUF Q8_0 everything resident | 46.57 GB | 2 h 37 min |
| Jetson AGX Orin 64GB NVIDIA | 64 GB | Barely usable | GGUF Q8_0 everything resident | 46.57 GB | 4 h 46 min |
| Radeon PRO W7900 AMD | 48 GB | Barely usable | GGUF Q8_0 everything resident | 46.57 GB | 4 h 51 min |
| GeForce RTX 4090 NVIDIA | 24 GB | Barely usable | BF16 sequential offload | 20.09 GB | 7 h 23 min |
| GeForce RTX 5090 Laptop NVIDIA | 24 GB | Barely usable | BF16 sequential offload | 20.09 GB | 9 h 22 min |
| GeForce RTX 5080 NVIDIA | 16 GB | Barely usable | GGUF Q5_K_M sequential offload | 15.16 GB | 10 h 49 min |
| GeForce RTX 4080 SUPER NVIDIA | 16 GB | Barely usable | GGUF Q5_K_M sequential offload | 15.16 GB | 11 h 42 min |
| GeForce RTX 4090 Laptop NVIDIA | 16 GB | Barely usable | GGUF Q5_K_M sequential offload | 15.16 GB | 12 h 11 min |
| GeForce RTX 4080 NVIDIA | 16 GB | Barely usable | GGUF Q5_K_M sequential offload | 15.16 GB | 12 h 29 min |
| GeForce RTX 5080 Laptop NVIDIA | 16 GB | Barely usable | GGUF Q5_K_M sequential offload | 15.16 GB | 12 h 49 min |
| GeForce RTX 5070 Ti NVIDIA | 16 GB | Barely usable | GGUF Q5_K_M sequential offload | 15.16 GB | 13 h 46 min |
| GeForce RTX 4070 Ti SUPER NVIDIA | 16 GB | Barely usable | GGUF Q5_K_M sequential offload | 15.16 GB | 13 h 50 min |
| Ryzen AI Max+ 395 64GB AMD | 64 GB | Barely usable | GGUF Q8_0 everything resident | 46.57 GB | 15 h 3 min |
| Ryzen AI Max+ 395 96GB AMD | 96 GB | Barely usable | GGUF Q8_0 everything resident | 46.57 GB | 15 h 3 min |
| Ryzen AI Max+ 395 128GB AMD | 128 GB | Barely usable | BF16 everything resident | 76.48 GB | 15 h 3 min |
| GeForce RTX 3090 Ti NVIDIA | 24 GB | Barely usable | BF16 sequential offload | 20.09 GB | 15 h 13 min |
| GeForce RTX 3090 NVIDIA | 24 GB | Barely usable | BF16 sequential offload | 20.09 GB | 17 h 9 min |
| RTX A5000 NVIDIA | 24 GB | Barely usable | BF16 sequential offload | 20.09 GB | 21 h 57 min |
| Apple M3 Ultra 96GB Apple | 96 GB | Barely usable | GGUF Q8_0 everything resident | 46.07 GB | 23 h 4 min |
| Apple M3 Ultra 256GB Apple | 256 GB | Barely usable | BF16 everything resident | 75.98 GB | 23 h 4 min |
| Apple M3 Ultra 512GB Apple | 512 GB | Barely usable | BF16 everything resident | 75.98 GB | 23 h 4 min |
| Apple M2 Ultra 64GB Apple | 64 GB | Barely usable | GGUF Q8_0 everything resident | 46.07 GB | 23 h 45 min |
| Apple M2 Ultra 128GB Apple | 128 GB | Barely usable | BF16 everything resident | 75.98 GB | 23 h 45 min |
| Apple M2 Ultra 192GB Apple | 192 GB | Barely usable | BF16 everything resident | 75.98 GB | 23 h 45 min |
| GeForce RTX 5060 Ti 16GB NVIDIA | 16 GB | Barely usable | GGUF Q5_K_M sequential offload | 15.16 GB | 25 h 23 min |
| GeForce RTX 4060 Ti 16GB NVIDIA | 16 GB | Barely usable | GGUF Q5_K_M sequential offload | 15.16 GB | 27 h 41 min |
| Jetson AGX Orin 32GB NVIDIA | 32 GB | Barely usable | BF16 sequential offload | 20.09 GB | 28 h 40 min |
| Radeon RX 7900 XTX AMD | 24 GB | Barely usable | BF16 sequential offload | 20.09 GB | 28 h 53 min |
| Apple M1 Ultra 64GB Apple | 64 GB | Barely usable | GGUF Q8_0 everything resident | 46.07 GB | 30 h 46 min |
| Apple M1 Ultra 128GB Apple | 128 GB | Barely usable | BF16 everything resident | 75.98 GB | 30 h 46 min |
| RTX A4000 NVIDIA | 16 GB | Barely usable | GGUF Q5_K_M sequential offload | 15.16 GB | 31 h 38 min |
| Radeon RX 7900 XT AMD | 20 GB | Barely usable | GGUF Q8_0 sequential offload | 16.5 GB | 34 h 30 min |
| Apple M4 Max 48GB Apple | 48 GB | Barely usable | GGUF Q5_K_M everything resident | 34.86 GB | 35 h 6 min |
| Apple M4 Max 64GB Apple | 64 GB | Barely usable | GGUF Q8_0 everything resident | 46.07 GB | 35 h 6 min |
| Apple M4 Max 128GB Apple | 128 GB | Barely usable | BF16 everything resident | 75.98 GB | 35 h 6 min |
| Radeon RX 9070 XT AMD | 16 GB | Barely usable | GGUF Q5_K_M sequential offload | 15.16 GB | 36 h 38 min |
| Radeon RX 7900 GRE AMD | 16 GB | Barely usable | GGUF Q5_K_M sequential offload | 15.16 GB | 38 h 37 min |
| Apple M3 Max 48GB Apple | 48 GB | Barely usable | GGUF Q5_K_M everything resident | 34.86 GB | 45 h 30 min |
| Apple M3 Max 64GB Apple | 64 GB | Barely usable | GGUF Q8_0 everything resident | 46.07 GB | 45 h 30 min |
| Apple M3 Max 96GB Apple | 96 GB | Barely usable | GGUF Q8_0 everything resident | 46.07 GB | 45 h 30 min |
| Apple M3 Max 128GB Apple | 128 GB | Barely usable | BF16 everything resident | 75.98 GB | 45 h 30 min |
Source: Wan-AI/Wan2.2-T2V-A14B-Diffusers . Downloaded 0.1 million times in the last month. See how we calculate, or browse every video model.