🧰 Prompt caching

The items that name it 2

• ALTK-Evolve: How learned memory tops the τ²-bench leaderboard on customer service tasks without a frontier model
  writeup2026-10-20

  Mentioned as a way to reclaim most of the extra input cost, since the guideline block is identical across every call.

• A model guide for the GPT-6 family
  article2026-10-02

  Reuses shared context to cut cost; cached input tokens cost up to 95% less than uncached, depending on the model