Tree of thoughts
Reasoning
Level 3 · Agentic & compositeExplore several different reasoning paths in parallel, evaluate them, then develop the best one.
Tree of thoughts extends chain of thought from a single reasoning line into a branching search: the model generates multiple different approaches, evaluates each one against explicit criteria, and only then commits to developing the most promising branch. This mirrors how people explore several options before deciding, rather than committing to the first idea.
When to use it
Use it for open-ended problems with more than one reasonable approach — strategy, creative problem-solving, design decisions — where comparing alternatives before committing genuinely improves the outcome.
When to avoid it
Avoid it for tasks with one clearly correct approach: generating and evaluating alternatives just adds tokens and latency without changing the outcome.
Advantages
- Reduces the risk of committing to a weak first idea
- Makes the model's evaluation criteria explicit and inspectable
- Well suited to genuinely open-ended, multi-option problems
Disadvantages
- Significantly more tokens than a single reasoning chain
- Needs clear evaluation criteria, or the comparison step is shallow
- Overkill for problems with only one reasonable approach
Examples
We need to reduce our app's onboarding drop-off rate. Input: 40% of new users abandon signup at the email verification step. Generate 3 independent approaches. Evaluate each against these criteria: implementation cost, expected impact, time to ship. Every approach must also respect these constraints: must not require a native app update (web-only fix). Then choose and develop the most promising one.
Pairs well with
Try it
Fill in the placeholders below and see the compiled prompt update live.
What you want to achieve.
The material to work from.
How many independent approaches to generate. Default: 3.
How should the alternatives be compared? (e.g. cost, impact, time to ship)
Rules, limits or things to avoid.
- 29
- Words
- 249
- Characters
- 0/5
- Placeholders filled
- 1 min
- Reading time
- Completeness0
- Clarity0
- Context0
- Specificity0
- Constraints0
- Format0
- Output0
- Constraints note100
- Success criteria0
- Scope focus100
- Verification step0
- Negative example0
- Confidence0
- Placeholder syntax100
- Markdown structure0
- Fill in the remaining placeholders.
- Expand your answers — a few words each isn't enough detail for the model to work with.
- Add background context so the model understands the situation.