In this video, Matthew Berman tests the capabilities of Llama 3.3, Meta’s latest text-only large language model, focusing on its reasoning and coding skills through practical examples. The tutorial demonstrates how to run Llama 3.3 locally and create a Python agent using Phidata to assess its performance on logic problems and coding challenges. Berman highlights the model’s strengths, such as its logical reasoning and code generation accuracy, while comparing it to other leading models in the AI landscape.