List your company

Text-to-Video Generation Inference on T2V Models HunyuanVideo, Wan2.2

199Latency (ms)RoME
-1,397.849,380.8320,159.530,938.17Apr 10, 2026

Evaluation results

MethodLinks
RoMEModel=Wan2.2, Parameters=1.3B, QKShape[BNSD]=[1, 12, 6630, 128]/272P2026.041993.9
RoPEModel=Wan2.2, Parameters=1.3B, QKShape[BNSD]=[1, 12, 6630, 128]/272P2026.04209—
RoMEModel=Wan2.2, Parameters=1.3B, QKShape[BNSD]=[1, 12, 17550, 128]/480P2026.047593.4
RoPEModel=Wan2.2, Parameters=1.3B, QKShape[BNSD]=[1, 12, 17550, 128]/480P2026.04790—
RoMEModel=Wan2.2, Parameters=14B, QKShape[BNSD]=[1, 40, 6630, 128]/272P2026.041,1204.3
RoPEModel=Wan2.2, Parameters=14B, QKShape[BNSD]=[1, 40, 6630, 128]/272P2026.041,170—
RoMEModel=Wan2.2, Parameters=14B, QKShape[BNSD]=[1, 40, 17550, 128]/480P2026.044,3503.3
RoPEModel=Wan2.2, Parameters=14B, QKShape[BNSD]=[1, 40, 17550, 128]/480P2026.044,500—
RoMEModel=HunyuanVideo, Parameters=13B, QKShape[BNSD]=[1, 24, 42840, 128]/540P2026.0414,4000.1
RoPEModel=HunyuanVideo, Parameters=13B, QKShape[BNSD]=[1, 24, 42840, 128]/540P2026.0414,410—
RoMEModel=HunyuanVideo, Parameters=13B, QKShape[BNSD]=[1, 24, 8192, 128]/720P2026.0440,1000.1
RoPEModel=HunyuanVideo, Parameters=13B, QKShape[BNSD]=[1, 24, 8192, 128]/720P2026.0440,120—