Reviews
A ledger of judgments over things experienced — played, read, watched, used, visited — one verdict and a small rubric each.
Models · 06
Claude截至 Fable 5——编码和长任务的天花板。ActiveCapability5 / 5Speed3 / 5Value3 / 5Honest about limits5 / 5GPT截至 5.6 Sol——编码和长任务的天花板仅次于 Fable,前端设计不如 Fable 5 和 K3。ActiveCapability5 / 5Speed4 / 5Value4 / 5Honest about limits4 / 5Kimi截至 K3——编码和长任务完全能胜任,前端天花板。ActiveCapability5 / 5Speed2 / 5Value4 / 5Honest about limits4 / 5DeepSeek截至 V4 Flash——编码和长任务完全能胜任,欠缺多模态能力,前端设计一般。ActiveCapability3 / 5Speed4 / 5Value5 / 5Honest about limits3 / 5Gemini截至 3.6 Flash——编码和长任务的可用水平,前端设计还行。ActiveCapability3 / 5Speed4 / 5Value3 / 5Honest about limits3 / 5GLM截至 GLM-5.2——编码和长任务完全能胜任,欠缺多模态能力,前端设计一般。ActiveCapability3 / 5Speed3 / 5Value4 / 5Honest about limits3 / 5