Claude Opus system prompts keep growing: Opus 5.5 hits 4,108 words, mostly refusal guidance
rajistics · x · 2026-09-23
Rajiv Shah updated his visual analysis of Claude's published system prompts, pushing back on the narrative that better models need shorter prompts.
Key data
- Opus 4 (May 2025): 1,714 words
- Opus 4.7 (Apr 2026): 3,686 words after safety, agentic behavior and product scaffolding expanded
- Opus 5.5 (released Sep 22): 4,108 words, up 883 words (+27.4%) from Opus 5's 3,225, with the increase mostly in refusal guidance
Takeaway: some prompt complexity does move into training, but the total system doesn't shrink — it shifts into safety policy, tool discovery, product context, memory, skills and app-specific harnesses. The curve went up, down, then up again; Opus 5.5 reversed the recent dip. The interesting question is where complexity moves, not whether prompts get shorter.
Related event: Opus 5.5 system prompt hits 4,108 words, defying 'shorter prompt' narrative(2 posts)→
More from coding & agent
- Google's MCP Apps extension lets MCP servers return interactive UI components — rseroter · 2026-09-23
- Dev runs 32 coding agents burning 5-10B tokens a day in Zed — generativist · 2026-09-23
- Stanford paper: self-organizing agent team hits 66.7%, beats best single model's 48.8% — dair_ai · 2026-09-23
- Google's ToolGrad reverses tool-use data generation, hits 83.1 BFCL with 12B Gemma-3 — dl_weekly · 2026-09-23
- Using AI agents to book things still costs as much mental effort as doing it yourself — nabeelqu · 2026-09-23
- LangChain to unveil new products Thursday, previewed live at Interrupt keynote — LangChain · 2026-09-23