List your company

Complexity and Throughput Analysis on Hardware Efficiency Benchmark H100 Configuration

93.8H100 Throughput (tokens/s)GDN-2 (matrix-gated)
22.0440.6759.377.93Jun 25, 2026

Evaluation results

MethodLinks
GDN-2 (matrix-gated)Hardware=Single H100, Model Scale=125M, Sequence length (T)=1024, Mini-batch (mb)=82026.0693.8————
CARVEHardware=Single H100, Model Scale=125M, Sequence length (T)=1024, Mini-batch (mb)=82026.0693.4————
Content-aware exact recurrentHardware=Single H100, Model Scale=125M, Sequence length (T)=1024, Mini-batch (mb)=82026.0624.8————
Delta RuleHardware=Single H100, Model Scale=125M, Sequence length (T)=1024, Mini-batch (mb)=82026.06—222—
Linear AttentionHardware=Single H100, Model Scale=125M, Sequence length (T)=1024, Mini-batch (mb)=82026.06—222—