Ultrathink | ClaudeLog
claudelog.com · 343 words · saved by 1 readers
Maximize test-time compute with ultrathink and revving mechanics to push Sonnet to its limits.
Ultrathink Unlock Claude's maximum reasoning power by adding "ultrathink" to your prompts. Re-introduced in v2.1.68, "ultrathink" triggers high effort mode for the next turn—combined with Plan Mode, it can push Sonnet to near-Opus performance on complex tasks. Ultrathink & Plan Mode I personally get outstanding results using ultrathink + Plan Mode with Claude Sonnet 4.5. This combination is especially useful when you don't want to reach for Claude 4 Opus. I have found it can often bridge the intelligence gap for complex tasks, and when that's not sufficient, I opt to rev the model by…
saved by
related reading
- A Guide to Claude Code 2.0 and getting better at using coding agents – sankalp's blogsankalp.bearblog.dev
- Prompting best practicesdocs.anthropic.com
- Best practices for Claude Code - Claude Code Docsanthropic.com
- How I Use Claude Code | Philipp Spiessspiess.dev
- Claude 4.5 Opus' Soul Document — LessWronglesswrong.com
- Claude's extended thinking \ Anthropicanthropic.com
- Claude Sonnet 4.5 System Cardassets.anthropic.com
- Claude Code Cheat Sheetcc.storyfox.cz
- Turn Claude From a Chatbot Into a Thinking Partner 🧠substack.com
- Claude Sonnet 4.5 is probably the “best coding model in the world” (at least for now)simonwillison.net
- Don’t Outsource Your Thinkingteltam.github.io
- The "think" tool: Enabling Claude to stop and think \ Anthropicanthropic.com