List your company

Inference Memory Analysis on NVIDIA H100 NVL (seq_len=1024)

744Model Weights SizeTernary Mamba
560.881,796.943,0334,269.06Jun 16, 2026

Evaluation results

MethodLinks
Ternary MambaConfiguration=Ternary (packed, proj.)2026.067441,100
FP16Configuration=FP162026.062,5992,932
Ternary MambaConfiguration=Ternary (sim.)2026.065,3226,932