Prompting Claude Opus 5.5
Anthropic published prompting guidance for Claude Opus 5.5, covering effort, agents, refusals, and visual inputs.
Anthropic's prompting guide for Claude Opus 5.5 says existing Claude Opus 5 prompts should still work, while the newer model produces output tokens more than 30 percent faster and often finishes tasks with fewer tokens. Effort defaults to medium, and Anthropic reports that medium effort matches or beats Claude Opus 5 at high effort on coding and knowledge-work evaluations. The guide covers effort calibration, always-on thinking, unattended and multiagent runs, safeguard refusals, progress updates, multi-app context, pasted text, and visual inputs. It advises setting max_tokens high enough to include thinking tokens, up to 128,000 for long agentic coding turns.
- Claude Opus 5.5 generates tokens over 30 percent faster than Opus 5.
- Medium effort can match Opus 5 at high effort on coding evaluations.
- Thinking is always on and counts toward the max_tokens limit.
- The guide covers agents, refusals, visuals, and pasted user text.
- Vendors
- Anthropic
- Products
- Claude API
- AI models
- Claude Opus 5.5Claude Opus 5
From the original article
14 min read at platform.claude.comBehavioral differences from Claude Opus 5 and the prompting and harness patterns that address them: effort calibration, thinking behavior in API integrations and chat, progress updates, unattended and multiagent tasks, safeguard refusals, frontend design, complex visual inputs, multi-app workflows, and pasted text in user messages. This guide covers the prompting patterns specific to Claude Opus 5. 5.
By Michelangelo11 · platform.claude.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.