Wan 2.2 A14B (T2V / I2V) VRAM Requirements
Wan 2.2 A14B (T2V / I2V) needs about 18.8 GB of VRAM with FP8 weights at 832×480 and 81 frames, loading one model at a time. In FP16 it needs 31.9 GB (42.7 GB with everything loaded), and GGUF Q4 brings it to 13.2 GB. The smallest consumer GPU that runs it is the RTX 4060 Ti 16GB with Q4 weights, or the Radeon RX 7900 XT 20GB in FP8.
18.8 GB peak VRAM
Wan 2.2 A14B (T2V / I2V) at 832×480, 81 frames with FP8 weights, loading one model at a time (ComfyUI default on smaller cards). Keeping everything loaded needs 24.4 GB.
- Diffusion model 13.0 GB
- Text encoders 5.3 GB
- VAE 0.2 GB
- Peak activations 5.0 GB
- Overhead 0.8 GB
GPU fit
- RTX 4060 8GB40% on GPU
- RTX 3060 12GB60% on GPU
- RTX 4060 Ti 16GB80% on GPU
- RTX 4090 24GBFitswith swapping
- RTX 5090 32GBFitsall loaded
- RTX 6000 Ada 48GBFitsall loaded
- RTX PRO 6000 Blackwell 96GBFitsall loaded
- Apple M4 Max 128GBFitsall loaded
Wan 2.2 A14B (T2V / I2V) VRAM by precision
At 832×480, 81 frames, tiled VAE. Text encoders in FP8 except for the FP16 row.
| Weights | Diffusion model | Peak (swapping) | All loaded | Smallest tier |
|---|---|---|---|---|
| FP16 / BF16 | 26.1 GB | 31.9 GB | 42.7 GB | 48 GB |
| FP8 | 13.0 GB | 18.8 GB | 24.4 GB | 24 GB |
| GGUF Q8_0 | 13.9 GB | 19.7 GB | 25.2 GB | 24 GB |
| GGUF Q5_K_S | 9.0 GB | 14.8 GB | 20.3 GB | 16 GB |
| GGUF Q4_K_S / NF4 | 7.4 GB | 13.2 GB | 18.8 GB | 16 GB |
VRAM by resolution
FP8 weights, with model swapping.
| Resolution | Latent tokens | Activations | Peak VRAM |
|---|---|---|---|
| 832×480, 81 frames | 32,760 | 5.0 GB | 18.8 GB |
| 1280×720, 81 frames | 75,600 | 11.5 GB | 25.4 GB |
About Wan 2.2 A14B (T2V / I2V)
Two 14B experts, one for high-noise steps and one for low-noise steps. Only one is used at a time, so peak memory is that of a single 14B model, but you need disk and system RAM for both. 81 frames is 5 seconds at 16 fps.
| Component | Size |
|---|---|
| Diffusion model | 14B parameters (MoE diffusion transformer) |
| Text encoders | umT5-XXL encoder (5.7B) |
| VAE | 0.127B parameters |
| Classifier-free guidance | Yes (2 passes per step) |
| Released | 2025-07 |
| License | Apache 2.0 |
Wan 2.2 A14B (T2V / I2V) VRAM: common questions
How much VRAM does Wan 2.2 14B need for 720p?
Expect about 25 GB with FP8 weights and model swapping at 1280×720 and 81 frames. On 12 to 16 GB cards, use GGUF Q4 to Q6 and block swapping.
Does Wan 2.2 A14B need 28B of VRAM because it has two experts?
No. ComfyUI loads the high-noise expert, then swaps to the low-noise one. Peak VRAM is one 14B model plus activations.
How much VRAM does Wan 2.2 A14B (T2V / I2V) need?
At its default 832×480 and 81 frames: about 31.9 GB in FP16 with model swapping (42.7 GB with everything loaded), 18.8 GB in FP8 and 13.2 GB with GGUF Q4 weights.