Unverified50% confidenceBenchmarkExact time
在SWE-Bench Pro基准上，Sonnet 5明显落后于Opus 4.8
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
7/22/2026
First Seen
Valid until: 10/20/2026
Sources
Related Claims
Unverified在SWE Bench Pro编程基准测试中，Sonnet 5取得63.2%，Opus 4.8取得69.2%80% 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80% similarUnverified在 Cognition 推出的第三方基准 Frontier Code 上，Sonnet 5 超越了 Opus 4.878% similarUnverifiedAnthropic可能直接跳过了Sonnet 4.7，从4.6直接升级到4.878% similarPartially Verified官方称Sonnet 5在推理、工具使用、编码和知识工作方面相比Sonnet 4.6均有显著提升，整体性能已逼近OPUS 4.8但价格更低77% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/593948API
curl https://kongchang.com/api/v1/knowledge/claims/593948MCP
get_claim(id=593948)