跳到正文
Vercel Blog(vercel.com)· Rohan Taneja·· 2026-09-22精選品質評分78

Claude Opus 5.5 在 AI Gateway 上可用

Claude Opus 5.5 now available on AI Gateway

重點導讀

Anthropic 的 Claude Opus 5.5 已在 AI Gateway 上上線,支援 1M-token 上下文視窗和 128K 輸出 token。

與 Opus 5 相比,速度提升約 30%,成本降低約 40%。

AI 伴隨問答

自備 Key 深度分析

針對這篇報導有疑問?可直接點選快捷問題或向 AI 追問背景脈絡與技術細節:

推薦理由

Opus 5.5 在速度和成本上相較前代提升顯著,適合需要大上下文和長任務的AI 助理工作,其 1M-token 上下文視窗和 128K 輸出令長任務更易監督,且支援自動工具選擇,提升開發效率。對開發者而言,Opus 5.5 的速度與成本優勢可顯著降低專案迭代週期。

正文 · 原文

Claude Opus 5.5 from Anthropic is now available on AI Gateway as anthropic/claude-opus-5.5.

Opus 5.5 is Anthropic’s strongest Opus model for agentic coding, long-running agent tasks, and professional knowledge work. It communicates progress, findings, and next steps clearly, making extended work easier to supervise.

It has a 1M-token context window and supports up to 128K output tokens. Compared with Opus 5, Anthropic has stated that Opus 5.5 performs at a similar level to Fable 5.1, but is ~30% faster and ~40% cheaper per task than Opus 5.

The new model includes two API changes that can turn previously valid requests into HTTP 400 errors:

  • Thinking is always adaptive. Requests that disable thinking or set a fixed thinking budget are rejected. Use an effort level and prompting to control reasoning depth.

  • Forced tool choice is unsupported. Requests cannot require a tool call or force a specific tool. Use automatic tool selection and prompt the model when it should call a tool. If you previously forced a tool call to return JSON, use structured outputs instead.

To use Claude Opus 5.5:

For lower latency, enable fast mode with speed: 'fast' or use anthropic/claude-opus-5.5-fast. In Claude Code, run /fast to toggle it for the session.

Opus 5.5 also supports regional inference and Zero Data Retention through AI Gateway. Try it in the model playground.

Read more

來源: Vercel Blog (vercel.com)
開啟原文

讀者討論

理性交流 · 即時討論