Reasoning Models: When AI Thinks Before It Speaks

End-of-lesson check

15 questions · take it digitally for instant feedback at tendril.neural-forge.io/learn/quiz/end-builders-models-reasoning-models-r8a8-teen

1. What do reasoning models do before generating their final answer?

   1. They ask the user clarifying questions first
   2. They spend extra compute time generating internal reasoning tokens
   3. They search the internet for the correct answer
   4. They simply respond faster than regular models

2. A student needs help writing a quick reply to a friend's text message. Which approach would be most appropriate?

   1. Use a regular (non-reasoning) AI model
   2. Wait for a reasoning model to finish its extended thinking
   3. Use a reasoning model like o3 for accuracy
   4. Use DeepSeek-R1 for faster processing

3. Which of the following problems would benefit MOST from using a reasoning model?

   1. Translating a sentence from English to Spanish
   2. Writing a friendly birthday greeting
   3. Summarizing a paragraph into three bullet points
   4. A complex math competition problem requiring multi-step logic

4. What is 'extended thinking' a feature of?

   1. Claude only
   2. ChatGPT only
   3. DeepSeek-R1 only
   4. Both Claude and other reasoning models

5. A developer is working on a Codeforces-level coding challenge. Which model would be particularly suitable for this task?

   1. DeepSeek-R1 or o3
   2. A speech recognition model
   3. A text-to-image generator
   4. A simple chatbot for customer service

6. What is the main trade-off when using a reasoning model?

   1. Faster responses but lower accuracy
   2. Cheaper cost but less capable reasoning
   3. Better for images but worse for text
   4. Higher accuracy but higher cost and longer response time

7. Which company developed the o3 reasoning model?

   1. OpenAI
   2. Meta
   3. Anthropic
   4. Google

8. What type of task would be a waste of a reasoning model's capabilities?

   1. Complex code debugging
   2. A simple email reply to a colleague
   3. Research synthesis comparing 20 academic papers
   4. A tough logic puzzle

9. What makes reasoning models particularly good at math and logic problems?

   1. They are trained only on math data
   2. They can spend extra compute on step-by-step reasoning before answering
   3. They randomly guess until they get the right answer
   4. They have access to a calculator built-in

10. Why might a reasoning model take longer to respond than a regular model?

    1. The internet connection is slower
    2. The servers are overloaded
    3. The model is generating additional reasoning tokens before answering
    4. The model is waiting for user input

11. What is DeepSeek-R1 an example of?

    1. A regular chatbot model
    2. An image generation model
    3. A reasoning model
    4. A speech-to-text model

12. When would it make sense to use a reasoning model for research synthesis?

    1. When synthesizing findings from 20 papers with citations
    2. When translating a paper to another language
    3. When looking up a single fact
    4. When comparing exactly 2 papers

13. What should guide your decision to use a reasoning model?

    1. Use them only for creative writing tasks
    2. Only use them when the answer matters more than speed
    3. Never use them because they are too expensive
    4. Always use them because they are the newest technology

14. What happens to the internal reasoning tokens that reasoning models generate?

    1. They are deleted immediately after use
    2. They are never shown to the user
    3. They may or may not be visible to the user
    4. They are always shown to the user

15. What is a key reason NOT to use reasoning models for every task?

    1. They require special hardware
    2. They are often wrong on simple tasks
    3. They cost significantly more and take longer
    4. They cannot handle simple tasks