🎯 Disaggregated prefill/decode serving

The items that name it 3

• Cross-request KV cache exposure in the GLM-5.3 Workers AI model (155 responses)
  writeup2026-10-26
• Long Live the Short King: Why 4-hi HBM Wins
  article2026-09-13

  Separates prefill and decode into different GPU pools.

• TPU Inference Externalization Full Steam Ahead - InferenceX
  article2026-09-07

  Separates prefill and decode; not yet fully optimised on the external TPU stack.