Users recently tested various large language model programming tools, including Codex, Claude Code, OpenCode T Rae, and China’s Minimax. Codex tends to exhaust resources quickly; third-party tools are unstable but fast; Claude Code is reliable but costly, with unexpected charges on private nodes; Chinese tools like Minimax respond quickly but have lower intelligence, frequently modifying code and causing issues. These experiences reveal differences in performance, cost, and stability among tools, providing practical reference for developers choosing suitable tools.
