ReAct vs Plan-and-Execute
A ReAct agent decides one action at a time, reasoning after every observation about what to do next. A plan-and-execute agent writes a complete multi-step plan first and then carries out the steps, which uses fewer tokens and allows parallel steps but adapts less well unless it replans when something goes wrong.
Definitions
- ReAct PTL-0058
- ReAct is a prompting pattern that interleaves reasoning traces ("Thought") with actions such as tool calls ("Action") and their results ("Observation"), letting a model plan, act, and update its plan in a loop.
- Plan-and-Execute PTL-0141
- Plan-and-execute is an agent design in which a planner model first writes a complete multi-step plan and executor calls then carry out the steps, rather than deciding one action at a time after every observation as in ReAct.
Differences
| ReAct | Plan-and-Execute | |
|---|---|---|
| When it plans | Before every action | Once up front (with replanning on failure) |
| Adaptability | High; each step sees the latest result | Lower; the plan may go stale |
| Cost and latency | Higher; reasons over a growing history each step | Lower; steps can run in parallel |
Which to use
- Use ReAct when each result determines the next step, as in open-ended research or debugging.
- Use plan-and-execute when the steps can be known in advance and speed or cost matter.
Sources
- Yao et al. (2022). ReAct: Synergizing Reasoning and Acting in Language Models.
- Xu et al. (2023). ReWOO: Decoupling Reasoning from Observations for Efficient Augmented Language Models.
- Kim et al. (2023). An LLM Compiler for Parallel Function Calling.
Cite this page
Protologue. (2026). ReAct vs Plan-and-Execute. In Protologue: A Taxonomy of Prompting and LLM Techniques (v1.4.0). https://protologue.com/compare/react-vs-plan-and-execute/