Claude Outperforms Deepseek in Coding and Analysis
Tom's Guide report compares Claude and Deepseek across 7 tasks
Translated by Qiyas from الراي — Original source
The website Tom’s Guide conducted a comparative test between Anthropic’s Claude 3.5 model and the Chinese Deepseek R1 model across seven practical tasks: programming, translation, financial analysis, literary creativity, mathematical problem-solving, long-context understanding, and Arabic language processing. The results showed that Claude outperformed Deepseek in four tasks (programming, financial analysis, mathematical problem-solving, and long-context understanding), while Deepseek excelled in two tasks (translation and Arabic language), with both models tying in literary creativity.
The technology editor noted that Deepseek demonstrated a deeper understanding of Arabic dialectal variations and the ability to generate natural-sounding Arabic text, whereas Claude was more accurate in Python programming and tabular analysis. The test revealed that a Chinese model outperformed an American one in Arabic, a finding that analysts described as a “stunning surprise.”
- In the programming task (building a simple app), Claude achieved 92% accuracy compared to Deepseek’s 68%.
- In translating a complex political article from English to Arabic, 85% of Arabic speakers preferred Deepseek’s translation due to its smoother style.
Observers note that this competition indicates that superiority is no longer the exclusive domain of American companies, as Chinese models are advancing rapidly. The report recommended using Claude for professional developers and Deepseek for those who work intensively with Arabic content.
