Anthropic API users beware: cache 5-minute window can make bundled credits hilariously wasteful
teortaxesTex · x · 2026-11-08
User teortaxesTex realized his use of Anthropic's bundled API credits was "hilariously wasteful" because his request reuse never lands within the prompt cache's 5-minute window.
Console screenshots show Claude Opus 5.5 burning 2.6M input tokens with only a 0.1% cache-read ratio — most traffic is 5-minute cache writes billed at full price, with write amortization near 0.00x. He quips that Anthropic should maybe distill something from DeepSeek for once.
More from Infra
- Shruti open-weights speech-to-text covers all 22 Indian languages offline in under 30MB — kingslyj · 2026-11-08
- Dev runs SmolLM-135M inference entirely in SystemVerilog via Verilator — cneuralnetwork · 2026-11-08
- How Robert Dennard's 1966 One-Transistor Idea Became the Foundation of Modern DRAM — cneuralnetwork · 2026-11-08
- Inkling 2's KV cache is 10x DeepSeek's, fueling doubts about RSI-driven architecture gains — bookwormengr · 2026-11-08
- Harbor turns one: 200T tokens/week, 10M+ Terminal-Bench runs, 1B+ rollouts — BenBlaiszik · 2026-11-08
- LLM-leaderboard-style atlas benchmarks silicon photonics foundries on component specs — jwt0625 · 2026-11-08