List your company

LLM Inference Efficiency on Uniform Sampled Documents

191.7TTFT (ms)Lazy-Attn
33.3481,102.2242,171.13,239.976Jun 3, 2026

Evaluation results

MethodLinks
Lazy-AttnHardware=H100 (High BW)2026.06191.76.2
CacheBlendHardware=H100 (High BW)2026.06274.84.3
Lazy-AttnHardware=A100 (Mid BW)2026.06372.56.9
CacheBlendHardware=A100 (Mid BW)2026.06565.34.5
Lazy-AttnHardware=A40 (Low BW)2026.06675.26.1
CacheBlendHardware=A40 (Low BW)2026.061,150.63.6
Prefix CachingHardware=H100 (High BW)2026.061,196.81
Prefix CachingHardware=A100 (Mid BW)2026.062,580.41
Prefix CachingHardware=A40 (Low BW)2026.064,150.51