Mistral Large and Codestral Quiz

Answers and explanations appear as you take the quiz.

1. What is Codestral specifically designed for?
   • Code generation and completion
   • Image generation
   • Speech recognition
   • Video editing
2. Where is Mistral AI headquartered?
   • San Francisco, USA
   • Paris, France
   • London, UK
   • Berlin, Germany
3. What architecture does Mixtral use to stay efficient?
   • A single dense network
   • Pure rule-based logic
   • Sparse mixture-of-experts
   • A recurrent neural net only
4. What does 'fill-in-the-middle' allow Codestral to do?
   • Translate spoken audio
   • Render 3D graphics
   • Compress files
   • Insert code using both the text before and after the cursor
5. How should developers interpret the licenses of Mistral downloadable weights and hosted services?
   • Check the exact model and release terms; weight licenses and hosted service terms are distinct
   • Assume every model and hosted endpoint shares Apache 2.0 terms
   • Assume downloading weights places them in the public domain
   • Assume a model weight license automatically permits reselling Mistral’s hosted API
6. Which capability makes Mistral Large useful for building agents?
   • Built-in printer drivers
   • Function/tool calling
   • Native blockchain mining
   • Analog signal processing
7. How many parameters are active per token in Mistral Large 3’s sparse mixture-of-experts model?
   • 3 billion active parameters
   • 22 billion active parameters
   • 41 billion active parameters
   • 675 billion active parameters
8. How many programming languages did Mistral say its original May 2024 Codestral model was trained on?
   • Exactly 8 languages
   • Only Python and JavaScript
   • More than 800 languages
   • More than 80 programming languages