ResearchBenchmarksLatency Measurement on LLaMA linear layers (11008 × 4096) 7BFollow0.0512Latency (ms)MBOK-0.073210.7663551.605922.445485May 28, 2025Evaluation ResultsMethodMethodLinksLatency (ms)Speed-upMBOKWEIGHT SIZE=11008 × 40...WEIGHT SIZE=11008 × 4096, Hardware=A100 GPU2025.050.05125.41FP16WEIGHT SIZE=11008 × 40...WEIGHT SIZE=11008 × 4096, Hardware=A100 GPU2025.050.2766—QUIP#WEIGHT SIZE=11008 × 40...WEIGHT SIZE=11008 × 4096, Hardware=A100 GPU2025.050.55990.49QTIPWEIGHT SIZE=11008 × 40...WEIGHT SIZE=11008 × 4096, Hardware=A100 GPU2025.053.16070.09