22,022 API calls analyzed: Fable 5.1 burns ~31% more tokens, plus 3 quota traps
tenequm · reddit · 2026-09-16
Based on 22,022 of their own Claude Code API calls, a Reddit user breaks down three reasons Claude quota drains faster this week:
- Never cleanly communicated: the cache pricing discount only applies to API usage — Boris Cherny's price-cut announcement explicitly listed "Enterprise, API, and SDK customers"; subscriptions are not included.
- The announced one: the temporary +50% weekly boost became a permanent +25% from Sep 13, i.e. 17% less quota.
- Open bug: messages sent to finished subagents miss the cache entirely, so their full context gets re-billed on every request — examples show only 16K of 243K/399K written tokens read from cache.
Their companion analysis found Fable 5.1 consumes 31% more tokens per prompt than 5.0.
More from coding & agent
- Microsoft's MemSteer turns unlabeled user histories into agent memory supervision — MicrosoftResearch · 2026-11-02
- She fired herself from her own Instagram: AI agent runs the account, views up 82.3% — heyshrutimishra · 2026-11-02
- Sourcegraph: AI writes code faster than teams can review it, driving more outages — AI Engineer · 2026-11-02
- Tsinghua & Microsoft's HarnessMM cuts agent token use by up to 87.93% while raising success rate — rohanpaul_ai · 2026-11-02
- GitHub launches Agent HQ: Anthropic, OpenAI, Google coding agents in Copilot — emmanuelvivier · 2026-11-02
- DeepSeek open-sources Super-Agent, a coding agent that reads team chats to minimize human input — emmanuelvivier · 2026-11-02