You do not need High or Max thinking effort for every AI request. Abdul Osama (عبدول أسامة) explains that a simple HTML text edit and a difficult programming problem call for different settings; using the highest level by default can be unnecessary.
“It’s not normal to leave it on Max all the time.” — Abdul Osama
Match the setting to the task
Claude is an AI assistant with different models and settings. ChatGPT is another AI assistant that can be used with different thinking levels. By effort, Abdul means how much processing the model is allowed to devote to a response—not an upgrade to its underlying abilities.
| Task | Abdul’s suggested approach | |---|---| | Editing text in HTML | Use a lighter model such as Sonnet at Low | | Routine work without much reasoning | Do not default to Max; Medium or Low may suffice | | Debugging or building infrastructure with complex logic | Choose a model suited to the task, and consider High if it is especially difficult |
Sonnet is a Claude model family. Other model names in the speech are unclear, so this page does not rely on their names or versions to make a recommendation.
“You have to judge how much thinking the task you’re asking for needs.” — Abdul Osama
Background
Tokens are units used to process text. Allowing more thinking can use more tokens, although actual usage depends on the task and settings. A higher setting does not guarantee a better answer every time.
What to do next
Decide whether your request is a straightforward edit or a problem requiring several reasoning steps. Start with an appropriate model and effort level, then increase the effort if the response is not sufficient—rather than keeping everything at the maximum.
Claude
ChatGPT
