Claude Code’s prompt cut is closer to 70%, and only on frontier models
PawelHuryn · x · 2026-07-22
The author says the widely cited "80% prompt cut" for Claude Code is overstated, and the real reduction is closer to 70% for frontier models.
What changed
- For Opus, the prompt shrinks from 2,686 words to 514 with memory off, but the memory rules were not deleted.
- When memory is included, the real drop is from 2,686 words to 830.
- The stripped prompt only applies to frontier models like Opus 4.8 and Fable 5; Sonnet 5 and Haiku 4.5 still keep the older, longer instruction set.
Why Anthropic did it
Anthropic’s explanation is simple: smarter models need fewer instructions.
The takeaway is that if your CLAUDE.md keeps growing, you may be writing for a weaker model than the one you are actually using.
Related event: Reverse Engineering Shows Claude Code Prompts Reduced by 70%(2 posts)→
More from coding & agent
- LangChain ships Managed Deep Agents 0.9, walkthrough from Interrupt London keynote — LangChain · 2026-10-26
- Spring AI 2.1.0-RC1 adds tool-call confirmation, pausing agent loops before real-data access — therealdanvega · 2026-10-26
- Morgan Stanley and US banks block third-party AI agents over credentials — thione · 2026-10-26
- METR red-teams OpenAI's agent monitoring, finds bypasses and gaps — thione · 2026-10-26
- OpenAI: coding agents weakened safety monitoring, one lied about it — thione · 2026-10-26
- ConvFill: A 0.6B Talker Buys Time While a Large Reasoner Whispers the Real Answer — andimarafioti · 2026-10-26