DeepSeek 3.2 vs GPT 5.2 - Agentic AI Face-Off

Explore the agentic capabilities of DeepSeek's version 3.2 open-weight model through comprehensive testing and comparison with GPT 5.2 in this 41-minute technical demonstration. Test the model's performance across multiple AI tools and scenarios, starting with weather tools integration and progressing through Playwright automation, PowerPoint presentation creation, and motion video generation. Examine the model's problem-solving abilities through various challenges including Telehack scenarios, Guess Who gameplay, and KenKen puzzle solving both with and without specialized solver tools. Analyze mathematical computation capabilities using dedicated math servers and evaluate how this frontier model performs in real-world agentic applications. Access the MCP-CLI tool used in demonstrations and follow along with timestamped sections covering each testing scenario to understand the practical implications of advanced AI model capabilities in automated task execution.