Anthropic 推出 Claude Sonnet 4.5，这是一款用于编码和自主任务的先进人工智能，性能优于竞争对手，无需人工输入即可运行长达 30 小时。 Anthropic launches Claude Sonnet 4.5, an advanced AI for coding and autonomous tasks, outperforming rivals and running up to 30 hours without human input.

人类已经发动了Claude Sonnet 4.5, 声称它是最先进的用于编码和自主任务执行的人工智能,能够在没有人类干预的情况下运行长达30小时,比其前身长4倍多。 Anthropic has launched Claude Sonnet 4.5, claiming it is the most advanced AI for coding and autonomous task execution, capable of running for up to 30 hours without human intervention—more than four times longer than its predecessor. 该模型在 " SWE-Bench " 验证和OSWorld等关键基准中取得了顶级得分,其业绩超过GPT-5和Gemini 2.5 Pro等竞争对手。 The model achieved top scores on key benchmarks like SWE-Bench Verified and OSWorld, outperforming rivals including GPT-5 and Gemini 2.5 Pro. 它在诸如网络搜索、日历更新和代码生成等现实世界任务中非常出色,拥有诸如Claude Agent SDK等新工具,并具有强化的编码功能。 It excels in real-world tasks such as web searches, calendar updates, and code generation, with new tools like the Claude Agent SDK and enhanced coding features. 该模型现已通过API和应用程序提供,价格不变,安全保护更强。 The model is now available via API and apps, with unchanged pricing and stronger safety protections.