DeepSeek-R1 has had a significant impact since its release. Cloud-based providers offer its use, and it can also be used locally with Ollama. In this article, the smallest DeepSeek-R1 model was tested using Ollama with several tasks.

• The notebook file is available on GitHub.
• Introduction: Tested with locally installed Jupyter and Ollama. Used the smallest model without GPUs. The notebook contains four test cases.

• Run the Code:

  • Code Generation and Explanation: Prompt led to Java code attempt instead of Python, and the code was incorrect and incomplete.
  • Mathematical Problem Solving: Correctly answered that two trains would meet at 10:47 AM.
  • Creative Writing With Constraints: Produced a short story within 200 words containing the required words.
  • Question Answering With Context: Answer had a mixture of correct and incorrect statements (Marie Curie's key discovery was polonium and radium, not the classified rays).
• Summary: Using Ollama with DeepSeek-R1, results were mixed. A larger model might have given better results, and other capabilities will be tested in future articles.