DeepSeek-V4.1-Flash on H20: 65.5% Lower Latency
Splitting prefill and decode across H20-3e GPUs cut DeepSeek-V4.1-Flash latency from 13.63 to 4.70 seconds and nearly tripled throughput.
Splitting prefill and decode across H20-3e GPUs cut DeepSeek-V4.1-Flash latency from 13.63 to 4.70 seconds and nearly tripled throughput.
Coding agents wrote one GPU program per model that runs the whole decode loop. On eight open models it beat the fastest general engine by up to 2.05x.