Dev Tools · 13h ago
GPT, Claude, DeepSeek face off on regex, refactoring, SQL
A developer benchmarked GPT, Claude, and DeepSeek on three coding tasks: CSV regex, nested-if refactoring, and SQL query generation. GPT and Claude tied on regex, Claude warned about performance; DeepSeek refused initially. On refactoring, all three produced clean code; on SQL, DeepSeek and Claude outperformed GPT on complex joins.
Meridian48 take
The test is small and subjective, but highlights how model refusal (DeepSeek) can be both a bug and a feature depending on the task.
Read the full reporting
GPT vs Claude vs DeepSeek на одних задачах: регулярка, рефакторинг, SQL →
DEV Community
llm-benchmarkcoding-ai