Unverified90% confidenceFactTime unknown
Anthropic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
UnverifiedAnthropic表示Sonnet 4.6在提示注入抵抗能力方面的表现与Opus 4.6相当74% similarUnverified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token73% similarUnverifiedSonnet 5 的 Max Effort 模式相比 Sonnet 4.6，每个 Intelligence Index 任务的输出 Token 多约 40%73% 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72% similarUnverified在SWE Bench Pro编程基准测试中，Sonnet 5取得63.2%，Opus 4.8取得69.2%70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/13802API
curl https://kongchang.com/api/v1/knowledge/claims/13802MCP
get_claim(id=13802)