I found that the current models are all specialized. So I basically use them in combination. Claude excels at deep reasoning, complex long-form writing, and coding. Gemini excels at multimodal tasks and images. ChatGPT excels at mathematics, logical reasoning, and creative writing. Grok excels at real-time X news. DeepSeek excels in Chinese scenarios. Perplexity excels at real-time search citations. Kimi excels at Chinese agents and has a large free quota.