Open-Source vs Closed AI: What Llama, Mistral, and DeepSeek Actually Mean

End-of-lesson check

15 questions · take it digitally for instant feedback at tendril.neural-forge.io/learn/quiz/end-builders-foundations-ai-open-source-vs-closed-r10a10-teen

1. Which statement best describes a closed AI model like GPT or Claude?

   1. Users can only access the model through the company's API service
   2. The model was created by Meta or Mistral
   3. The model can be run on any personal computer without internet
   4. The model weights are freely available for anyone to download

2. What company developed the Llama model?

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

3. What are 'model weights' in an AI system?

   1. The money paid to train an AI model
   2. Numerical parameters that store what the model has learned
   3. Safety rules that prevent harmful outputs
   4. The servers where AI models are hosted

4. What happened to Nvidia's market value after DeepSeek-V3 was released?

   1. It doubled in value
   2. It remained exactly the same
   3. It dropped by $1 trillion
   4. It increased by $1 trillion

5. What is a key advantage of running an open-source AI model locally on your own computer?

   1. It will cost less than using a paid API
   2. It can only be used for research purposes
   3. It doesn't require internet and your data stays private
   4. The model will always be more intelligent than cloud-based versions

6. Which of the following is an example of a closed AI model?

   1. GPT-4
   2. Llama 3
   3. DeepSeek-V3
   4. Mistral

7. Approximately how much did it cost to train DeepSeek-V3, and how does this compare to reported GPT-4 costs?

   1. $100 million, about the same as GPT-4
   2. $5 million, compared to GPT-4's $100 million+
   3. $500 million, much less than GPT-4
   4. $1 billion, about the same as GPT-4

8. What can you do with model weights after downloading an open-source model?

   1. Share them without any restrictions on usage
   2. Modify and run the model on your own hardware
   3. Convert them into a different company's model
   4. Use them only through the company's website

9. What does it mean for an open-source AI model to be 'uncensored'?

   1. The model cannot answer any questions
   2. It automatically generates only safe content
   3. It can only be used by adults
   4. It lacks safety guardrails that restrict certain outputs

10. Why might someone choose a closed AI model like GPT-4 over an open-source alternative?

    1. Closed models allow you to modify the model
    2. Closed models require no technical knowledge to create
    3. Closed models typically offer higher capability and easier access
    4. Closed models are always free to use

11. What hardware specification is mentioned for running Llama 3 8B on a personal Mac computer?

    1. A quantum computer
    2. A standard Windows laptop from 2015
    3. Any iPhone with internet access
    4. An M-series Mac with 16GB or more of RAM

12. What is the primary benefit of running an AI model locally rather than through a cloud API?

    1. Local models are free and never have usage limits
    2. Local models are always faster
    3. Your data stays on your device and doesn't require internet
    4. The local model will always produce better answers

13. Which company created the Claude model?

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

14. Which model was described in the lesson as the 'sweet spot' for free local models on consumer laptops in early 2025?

    1. DeepSeek-V3
    2. Llama 3 8B
    3. Mistral Small 3
    4. GPT-4

15. What is required to run Ollama and use a local AI model on your computer?

    1. Downloading and installing the Ollama software
    2. Access to a supercomputer
    3. An active internet connection at all times
    4. A paid subscription to OpenAI