DeepSeek V3.2 model has released its latest results in the Livebench benchmark, with a comprehensive comparison against industry-leading AI models such as Claude 4.5 Opus Thinking, Gemini 3 Pro Preview, GPT-5, and others. The test results show that V3.2 ranked ninth in reasoning tasks, sixteenth in programming capabilities, fourteenth in agent programming abilities, tenth in mathematical skills, and showed outstanding performance in data analysis, ranking third. These data points reflect the rapid iteration of current AI technology and intense competition among models, providing important references for AI practitioners, researchers, and developers to help evaluate the performance pros and cons of different models and promote the frontier development of artificial intelligence technology. The test results also highlight DeepSeek’s competitiveness in specific domains, particularly its strong performance in the field of data analysis.
