Glossary
Reasoning Models
LLMs that explicitly plan and reason before producing an answer.
Definition
Reasoning models are LLMs — or model modes — that perform an explicit internal reasoning step before returning a final answer. Examples include OpenAI's o-series, Claude with extended thinking, and Google's Gemini Thinking. They trade higher latency and cost for measurably better performance on hard problems.
Context
Reasoning models are best on well-defined hard problems: math, code, complex planning, agentic decisions. They are overkill for simple retrieval or short-form generation. In production, sophisticated systems route requests between reasoning and non-reasoning modes based on task difficulty to balance quality with cost and latency.