Verified65% confidenceFactExact time
Claude 3.5 Sonnet在代码生成基准测试HumanEval中达到了92%的通过率
3
Sources
65%
Confidence
Medium-term (~90 days)
Relevance
7/2/2026
First Seen
Valid until: 9/30/2026
Sources
Claude Opus/Sonnet/Haiku怎么选？实战对比与选型指南
bilibiliAI破局-老高6/8/2026
Related Claims
UnverifiedClaude 3.5 Sonnet在HumanEval基准测试上的得分超过90%，而2021年最初的Codex模型得分仅为28.8%79% similarUnverified在困难题中，Claude Sonnet 4.6拿下全场最高93分，注释最完整、测试最严谨75% similarUnverifiedClaude 3.5 Sonnet在SWE-bench软件工程基准测试上突破49%的解题率74% similarUnverified以 Claude 3.5 Sonnet 为代表的模型在代码生成的一次通过率（pass@1）上达到了工程实用水准74% similarVerifiedClaude 3.5 Sonnet在SWE-bench等代码基准测试中长期领先73% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/50047API
curl https://kongchang.com/api/v1/knowledge/claims/50047MCP
get_claim(id=50047)