List your company

Video Generation Inference Speed on Wan 768P Video (81 frames, 1280x768, without padding) 2.1

92Inference Time (s)OSP-Next
0.8616.41,2321,847.6May 27, 2026

Evaluation results

MethodLinks
OSP-NextHardware=Nvidia H200, Software Stack=torch.compile enabled, FlashAttention3, Device Count=8 GPU, Resolution=(81, 768, 1280), Padding=Without padding2026.05921.52
Wan2.1Hardware=Nvidia H200, Software Stack=torch.compile enabled, FlashAttention3, Device Count=8 GPU, Resolution=(81, 768, 1280), Padding=Without padding2026.05140—
OSP-NextHardware=Nvidia H200, Software Stack=torch.compile enabled, FlashAttention3, Device Count=4 GPU, Resolution=(81, 768, 1280), Padding=Without padding2026.051731.64
Wan2.1Hardware=Nvidia H200, Software Stack=torch.compile enabled, FlashAttention3, Device Count=4 GPU, Resolution=(81, 768, 1280), Padding=Without padding2026.05283—
OSP-NextHardware=Nvidia H200, Software Stack=torch.compile enabled, FlashAttention3, Device Count=2 GPU, Resolution=(81, 768, 1280), Padding=Without padding2026.053421.64
OSP-NextHardware=Ascend 910C, Software Stack=torch.compile disabled, SDPA, Device Count=4 NPU, Resolution=(81, 768, 1280), Padding=Without padding2026.053541.66
Wan2.1Hardware=Nvidia H200, Software Stack=torch.compile enabled, FlashAttention3, Device Count=2 GPU, Resolution=(81, 768, 1280), Padding=Without padding2026.05561—
Wan2.1Hardware=Ascend 910C, Software Stack=torch.compile disabled, SDPA, Device Count=4 NPU, Resolution=(81, 768, 1280), Padding=Without padding2026.05587—
OSP-Next-HiF8Hardware=Ascend 950PR, Software Stack=torch.compile disabled, SDPA, Device Count=1 NPU, Resolution=(81, 768, 1280), Padding=Without padding, Quantization=HiF82026.055902.27
OSP-NextHardware=Nvidia H200, Software Stack=torch.compile enabled, FlashAttention3, Device Count=1 GPU, Resolution=(81, 768, 1280), Padding=Without padding2026.056451.64
OSP-NextHardware=Ascend 910C, Software Stack=torch.compile disabled, SDPA, Device Count=2 NPU, Resolution=(81, 768, 1280), Padding=Without padding2026.056731.74
OSP-NextHardware=Ascend 950PR, Software Stack=torch.compile disabled, SDPA, Device Count=1 NPU, Resolution=(81, 768, 1280), Padding=Without padding2026.057621.76
Wan2.1Hardware=Nvidia H200, Software Stack=torch.compile enabled, FlashAttention3, Device Count=1 GPU, Resolution=(81, 768, 1280), Padding=Without padding2026.051,061—
Wan2.1Hardware=Ascend 910C, Software Stack=torch.compile disabled, SDPA, Device Count=2 NPU, Resolution=(81, 768, 1280), Padding=Without padding2026.051,168—
Wan2.1Hardware=Ascend 950PR, Software Stack=torch.compile disabled, SDPA, Device Count=1 NPU, Resolution=(81, 768, 1280), Padding=Without padding2026.051,338—
OSP-NextHardware=Ascend 910C, Software Stack=torch.compile disabled, SDPA, Device Count=1 NPU, Resolution=(81, 768, 1280), Padding=Without padding2026.051,3531.75
Wan2.1Hardware=Ascend 910C, Software Stack=torch.compile disabled, SDPA, Device Count=1 NPU, Resolution=(81, 768, 1280), Padding=Without padding2026.052,372—