Fireworks AI Achieves 1.9–2.4× Speedup for MiniMax M3 Sparse Attention on NVIDIA Blackwell GPUs

Fireworks built a KV-stationary sparse-attention kernel for MiniMax M3 on NVIDIA Blackwell (SM100), reaching ~980 TFLOP/s: 1.9–2.4× a query-stationary baseline and ~1.6× open-source MSA. The post walks through the Q-outer vs KV-outer design space, an I/O