Anthropic推出了Claude Sonnet 5.5模型，这是其Claude 5.5系列的最新中端模型。2相比前代Sonnet 5，新模型运行速度提升超过30%，成本降低30%。2该模型定价保持不变，输入令牌为每百万$2，输出令牌为每百万$10，缓存读取为每百万$0.20。2
Sonnet 5.5在多个性能基准上展现出明显进步。2在编程领域，该模型在FrontierCode高难度设置下相比Sonnet 5高出10分，同时成本仅为其1/15。2在Terminal-Bench 4.0评测中得分达70.6%，而Sonnet 5仅为10.3%。2在GDPval-AA评测中，Sonnet 5.5得分已接近Opus 5.5，超过Sonnet 5约400分。2此外，Anthropic宣称Sonnet 5.5在代理编码基准测试中表现优于Opus 5.5。1
在安全防护方面，Sonnet 5.5成为首个配备网络安全防护的Sonnet模型，其防护措施与Opus 5相当。2该模型也是首个加入防止推理提取安全分类器的Sonnet型号。2目前，Sonnet 5.5已在AWS、Google Cloud和Microsoft Azure等主流云平台上线。2Anthropic计划在未来数周内发布新版本的Haiku模型。1
Anthropic has released Claude Sonnet 5.5, a mid-tier artificial intelligence model positioned as a practical assistant for everyday work tasks.12 The new model delivers 30% faster performance compared to its predecessor Sonnet 5, while reducing operational costs by the same margin.2 Pricing remains unchanged at $2 per million input tokens and $10 per million output tokens, with cached reads at $0.20 per million tokens.2
The company emphasizes Sonnet 5.5's particular strength in code writing and office document generation.1 In specialized benchmarks, the model demonstrates notable improvements: it achieved a 70.6% score on Terminal-Bench 4.0, compared to Sonnet 5's 10.3%, and scored approximately 400 points higher than Sonnet 5 on GDPval-AA testing, approaching Opus 5.5's performance level.2 On FrontierCode programming evaluations at high difficulty settings, Sonnet 5.5 exceeded Sonnet 5 by 10 points while operating at just one-fifteenth of the cost.2 Sonnet 5.5 matches Opus 5's cybersecurity capabilities, marking the first Sonnet model to meet these enhanced security standards.12 It is also the first model in the Sonnet line to include a security classifier designed to prevent reasoning extraction.2
The model is now available across major cloud platforms including AWS, Google Cloud, and Microsoft Azure.2 Anthropic has announced plans to release an updated version of its Haiku model in the coming weeks.1
评论
还没有评论，欢迎留下第一条。