Suspected GPT-5.6 Cache Billing Anomaly
ClassicMain · reddit · 2026-07-11
A developer noticed a potential anomaly with prompt caching in the GPT-5.6 family:
- Using two identical requests on the original api.openai.com, cachewritetokens appear on the first request, but cachedtokens remains 0 for subsequent ones.
- This implies the system recognizes the cached prefix and avoids duplicate writes, but the read side isn't receiving the expected cache deduction.
- They noted that comparing GPT-5.5 within the same organization shows normal cache read billing.
- The post includes a minimal reproduction: two curl commands spaced 30 seconds apart reveal the issue.
The author concludes that if this isn't just a usage stats display bug, enabling caching on GPT-5.6 might be worse than not using it, prompting them to roll back to GPT-5.5 in their production routing for now.
More from Models
- Kimi K3.1 Pro tops MathArena open models, closing frontier gap to about 6 months — teortaxesTex · 2026-10-26
- Ruff creator Charlie Marsh: frontier models find infinite bugs in any software — charliermarsh · 2026-10-26
- Dev testing one skill across Opus 5.5 and Sol 6.1 finds instruction following 'all over the map' — lucasmeijer · 2026-10-26
- supermemory unveils learner-1, claiming continual learning for any model — julianweisser · 2026-10-26
- Anthropic says Opus 5.5 cuts cost per task by ~40% vs Opus 5 — taufiqintech · 2026-10-26
- AI search APIs ranked: top 12 within 7 points, yet prices differ nearly 7x — FinanceYF5 · 2026-10-26