Open-Source vs Proprietary LLMs for Business Quiz

Answers and explanations appear as you take the quiz.

1. What does the term open-weight most precisely mean?
   • The model is free to use with no license terms
   • The model's training data is fully public
   • The model can only run on open-source hardware
   • The trained parameters can be downloaded, but training data and often code are not released
2. Under Meta's Llama community license, which companies need a separate license from Meta?
   • Any company using the model commercially
   • Companies above 700 million monthly active users
   • Companies outside the United States
   • Companies with more than 100 employees
3. Which license was Mistral 7B released under?
   • Apache 2.0
   • A custom community license with a user threshold
   • A proprietary API-only license
   • No license at all
4. Which downside of proprietary models does the guide highlight?
   • They cannot be accessed through an API
   • They never offer enterprise support
   • The vendor can change prices, update behavior or retire model versions on its own schedule
   • They always use customer data for training
5. How should a business confirm whether a proprietary provider trains on its API data?
   • Assume it never does
   • Assume it always does
   • Ask the model itself
   • Check the contract and terms rather than assuming
6. Why does the guide recommend putting model calls behind a thin internal interface?
   • So switching providers means changing configuration rather than application code
   • To make models answer faster
   • To avoid paying for output tokens
   • Because licenses require it
7. Why should you pin model versions where possible?
   • Pinned versions are always cheaper
   • Pinning removes all licensing obligations
   • An unannounced change in behavior can break prompts tuned for an earlier version
   • Pinning lets you download proprietary weights
8. Which organization published an Open Source AI Definition in 2024?
   • Meta
   • The Open Source Initiative
   • The European Commission
   • NVIDIA