Users of Anthropic's Claude Code Pro Max plan are experiencing rapid quota exhaustion, with some reporting their 5x quota being depleted in as little as 1.5 hours. The issue appears to stem from how "cache_read" tokens are accounted for against rate limits, potentially counting at full rate instead of the expected reduced rate. This behavior negates the benefits of prompt caching and causes quota to deplete quickly, even with moderate usage, exacerbated by background sessions also consuming shared quota. AI
影响 Potential issues with quota management and token accounting could impact user experience and cost predictability for Claude Code users.
排序理由 This is a bug report about a specific product's quota system, not a new model release or significant industry event.
在 HN — claude-code stories 阅读 →
AI 生成摘要 · Google Gemini · 来自 1 个来源。 我们如何撰写摘要 →