Reads more as you go up
A few pages at Low, each school’s main books at High.
One setting decides how many pages the agent reads before it writes, and how fully it explains.
A few pages at Low, each school’s main books at High.
From key points to each view and its evidence.
Higher levels read deeper, never wider than you set.
Every citation still points to a page that was read.
One question at both levels, from recorded answers.
Each level tells the agent how much to read before it writes, and how fully to explain.
| Low | Medium | High | Extra high | Max | |
|---|---|---|---|---|---|
| Reads | A few key pages | One per school | Main books | Later works too | Same as Extra high |
| The direct answer, with cited points | |||||
| Each view with its evidence | |||||
| A short summary | |||||
| The terms of the question explained | |||||
| Replies to each view | |||||
| Where the views agree and differ | |||||
| The stronger view, per its sources | |||||
| A practical conclusion | |||||
| Background, and why the views differ | |||||
| Objections, replies and related issues | |||||
| The model’s longest reasoning |
Models without effort levels answer at Medium.
The Retry menu asks the question again, with the same model or another.
Medium for most questions. Low to check a single point. High or above when you are writing on the question, or the schools disagree.
Yes. It reads more pages and writes more, so it takes longer and uses more of your Claude or ChatGPT plan, or costs more with an API key. Extra high and Max are the slowest.
No. Search scope decides which books the agent can search. Depth decides how much of them it reads.
Yes. The model and level you pick are kept for each provider, so your next question uses them too.
Yes. Both offer Low, Medium, High, Extra high and Max. A model without levels answers at Medium.