May 31, 2026
Anthropic just added effort levels to Claude, and most users have no idea this exists yet.
Right next to the model selector in claude.ai, there is now an effort control that lets you choose how hard Claude thinks before responding.
Four levels: Low, Medium, High, and Max.
Higher effort means better answers, but slower responses and a faster rate limit consumption.
Lower effort means faster answers that use fewer of your daily limits.
If you have been using Claude on the default setting without touching this, you are either overpaying with your limits on simple tasks or leaving quality on the table for complex ones.
If I want a complex question answered, I discovered the best option is to use max
I tried max with this prompt :
I tested all four levels on different types of tasks to find out when each one makes sense.
Let me walk you through what I found.
Understanding the 4 Effort Levels
I like to compare this with what we have in Claude Code.
If you are a regular user, you already know we have a slash effort command that loads the effort levels:
But,
The effort control in Claude.ai sits in the model selector dropdown on claude.ai. Click the model name, then click "Effort" to expand the options.
Here is what each level does and when Anthropic recommends using it.
Low (Default on claude.ai)
Low effort gives you the fastest responses with minimal reasoning depth. Claude answers quickly without overthinking.
This is currently the default on claude.ai, which means every user who has never touched this setting is running on Low.
Best for:
Quick factual questions ("What is the capital of Kenya?")
Simple translations and formatting tasks
Brainstorming where you want fast ideas without deep analysis
Casual conversations where speed matters more than depth
Medium
Medium adds more reasoning without the full cost of deeper thinking. Claude takes a few extra seconds and produces more thorough responses.
Best for:
Email drafting and professional writing
Summarizing articles and documents
Explaining concepts at a moderate level of detail
Research questions that need some nuance
High
High is where Claude starts to think carefully about complex problems.
Responses take longer, but the quality difference is seen on anything that requires multi-step reasoning.
Best for:
Code generation and debugging
Detailed analysis of documents or data
Writing that requires careful structure and argumentation
Technical explanations that need accuracy
Max
Max pushes Claude to its deepest reasoning capability.
This is the most thorough setting and takes the longest to respond. It also consumes your rate limits the fastest.
The lock icon next to Max indicates it may require a higher-tier plan depending on your subscription.
Best for:
Complex coding tasks with multiple files and dependencies
Deep research requiring synthesis across many sources
Mathematical proofs and logical reasoning
Architecture decisions and system design
Thinking Toggle
Below the effort levels, there is a "Thinking" toggle that says "Can think for more complex tasks."
When enabled, Claude shows its reasoning process as it works through the problem. This pairs well with High and Max effort levels for tasks where you want to see how Claude arrived at its answer.
The combination of Max effort plus Thinking enabled gives you Claude's absolute best output, but expect slower responses and heavier rate limit usage.
Testing All 4 Levels on the Same Task
I gave all four effort levels the same prompt to see the difference in output quality and response time.
The test prompt:
Explain how JWT refresh token rotation works and write aPython implementationThis is a medium complexity task that involves both explanation and code generation. It should reveal clear differences across the levels.
These were my findings, side-by-side summary :
The jump from Low to Medium is easy to notice. The jump from Medium to High is significant. The jump from High to Max adds polish and completeness, but the core quality is already strong at High.
Effort Levels Combined with Model Selection
The effort control works across all three models: Opus 4.8, Sonnet 4.6, and Haiku 4.5.
This creates 12 possible combinations, and choosing the right one saves your limits.
Smart Combinations
Not every model needs every effort level. Here are the combinations that make sense:
Combinations You Should Avoid
Opus + Low:_ You are paying for the most capable model and then telling it not to think. This wastes your Opus rate limit allocation on responses that Haiku could produce at the same quality._
Haiku + Max:_ You are asking a model designed for speed to do deep reasoning. The output will not match what Sonnet or Opus produces at the same effort level. You hit diminishing returns quickly._
Any model + Max on simple tasks:_ If you ask Claude to translate a sentence or format a date string at Max effort, it will overthink a 2-second task and consume limits._
Claude Code Masterclass Course
Every day, I'm working hard to build the ultimate Claude Code course, which demonstrates how to create workflows that coordinate multiple agents for complex development tasks. It's due for release soon.
It will take what you have learned from this article to the next level of complete automation.
New features are added to Claude Code daily, and keeping up is tough.
The course explores Agents, Hooks, advanced workflows, and productivity techniques that many developers may not be aware of.
Once you join, you'll receive all the updates as new features are rolled out.
This course will cover:
Advanced subagent patterns and workflows
Production-ready hook configurations
MCP server integrations for external tools
Team collaboration strategies
Enterprise deployment patterns
Real-world case studies from my consulting work
If you're interested in getting notified when the Claude Code course launches, click here to join the early access list →
( Currently, I have 70,000+ already signed-up developers)
I'll share exclusive previews, early access pricing, and bonus materials with people on the list.
Let's Connect!
If you are new to my content, my name is Joe Njenga
Join thousands of other software engineers, AI engineers, and solopreneurs who read my content daily on Medium and on YouTube where I review the latest AI engineering tools and trends. If you are more curious about my projects and want to receive detailed guides and tutorials, join thousands of other AI enthusiasts in my weekly AI Software engineer newsletter
If you would like to connect directly, you can reach out here:
AI Automation Software Engineer (10+ Years Experience) My expertise spans LLMs, vector databases, RAG pipelines, MCP servers, and intelligent automation workflows that solve…
Follow me on Medium | YouTube Channel | X | LinkedIn | GitHub
By Freedium. — made with care.