What is a reasoning model and how does it differ from a standard LLM?
LLM & AILevel: FoundationWhat Are Reasoning Models
What is a reasoning model and how does it differ from a standard LLM?
Hints
- Identify the stage: pretraining, adaptation, generation or evaluation.
- Name the inputs, constraints and a case where the approach fails.
Answer
A standard LLM maps input → output in a single forward pass through the generation loop. A reasoning model is trained to produce an extended internal chain of thought — often called "thinking" — before generating its final answer.
Key differences:
| Aspect | Standard LLM | Reasoning Model |
|---|---|---|
| Output | Direct answer | Thinking trace + answer |
| Tokens generated | ~answer length | Thinking (100s–1000s) + answer |
| Latency | Fast | Slower (more tokens) |
| Accuracy on hard tasks | Lower | Significantly higher |
| Cost | Low | Higher (more tokens billed) |
| Best for | Simple Q&A, generation | Math, coding, multi-step logic |
Examples: OpenAI o1/o3, DeepSeek-R1, Claude 3.7 Sonnet (extended thinking), Gemini 2.0 Flash Thinking.
Source and licence: JustInCache & contributors — Awesome AI Interviews · Apache-2.0