Chain-of-thought is a prompting technique for breaking complex objectives into smaller, interpretable steps. Guide step-by-step reasoning within one prompt or divide the work across chained model calls. section 6.2.1
Reasoning models perform step-by-step reasoning internally, making explicit procedural prompts such as chain-of-thought, tree of thought, and self-consistency largely redundant. Focus prompts on the task, grounding data, constraints, examples, and tools; over-specifying internal steps can constrain the model’s trained reasoning. section 6.3.7




















