Claude Sonnet 5

嘿，猎人们！👋

今天我很兴奋来推荐 Claude Sonnet 5。

Claude Sonnet 5 是 Anthropic 至今最具智能体（agent）能力的 Sonnet——它能自主规划、使用浏览器和终端等工具，并独立完成复杂任务。在推理、编程和知识工作方面有大幅改进，性能接近 Opus 4.8，但成本更低。

现已面向所有 Claude 用户开放，包括 Free、Pro、Max、Team 和 Enterprise 版本。你会用它来做什么？

用了"小 Fabel"一段时间，昨晚开始上手的。说实话感觉跟 Opus 4.6 一样好用。🙏

I run Sonnet in long agent loops, so the thing I'm watching is tool-call stability: how many calls it chains before it drifts off the original plan. Older Sonnets would start second-guessing a decision they'd already committed to around call 15, which quietly derails an autonomous run. If Sonnet 5 holds the plan through a deep terminal-and-browser session, that's worth more to me than a few SWE-bench points. Anyone pushed it on long-horizon stability yet?

我有个做开发的朋友老是在对比不同 AI 模型的编程能力。我肯定要发给他，因为我知道他会想跑个 benchmark。

Sonnet 5 is doing something no other model does out of the box: spinning up its own sub-agents without being asked. pro tip: pair it with Opus for planning and let it execute the well-scoped work.

available right now on products like @Kilo Code and @v0 by Vercel.

几个月前从 ChatGPT 切到了 Claude，再也没回头。整个产品都构建在它上面，真正重要的是遇到问题崩溃时，它能帮你推理出艰难的决定。直接生成代码反而是容易的部分。我很好奇 Sonnet 5 的"智能体"能力提升在实际中能体现多少，而不仅仅是干净跑分的编程 benchmark——那才是真正干活的场景。