Unverified90% confidenceFactTime unknown
最大effort设置下Opus 4.5超越Sonnet 4.5达4.3个百分点，同时仍减少48%的token
1
Sources
90%
Confidence
Medium-term (~90 days)
Relevance
5/31/2026
First Seen
Valid until: 8/29/2026
Sources
Claude Opus 4.5工程测试碾压人类：AI编程能力全面超越顶尖工程师
bilibili枫叶边城
Related Entities
Related Claims
UnverifiedSonnet 5相较Opus 4.8 Max仅便宜约72%78% similarUnverified在SWE-Bench Pro基准上，Sonnet 5明显落后于Opus 4.874% similarUnverified实测者建议对于绝大多数任务继续使用Opus 4.8以获得更好效果，认为Sonnet 5令人失望74% similarUnverifiedAnthropic在API中引入了effort参数，中等effort可匹配Sonnet 4.5的最佳SWE-Bench分数同时减少76%的输出token73% similarUnverifiedSonnet 4.6与Opus 4.6相比，在大多数基准测试中表现几乎持平，但推理速度更快、调用成本更低72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/13803API
curl https://kongchang.com/api/v1/knowledge/claims/13803MCP
get_claim(id=13803)