Unverified50% confidenceBenchmarkExact time

早期GPT-3的HumanEval Pass@1不足10%，而GPT-4o、Claude 3.5 Sonnet等顶尖模型已稳定超过85%

1

Sources

50%

Confidence

Medium-term (~90 days)

Relevance

7/21/2026

First Seen

Valid until: 10/19/2026

Sources

Vibe Coding入门：零基础用AI造软件的新方式

Related Claims

UnverifiedGPT-4和Claude 3在HumanEval代码基准测试上的pass@1通过率超过85%81% similar VerifiedClaude 3.5 Sonnet在多项基准测试中与GPT-4o不相上下78% similar Partially Verified三个 GPT-5.6 模型在长程代理微调任务上均拿满分 10 分，而 GPT-5.5 和 Opus 4.7 在该任务只能拿两三分75% similar Unverified主流商业模型如GPT-4o、Claude 3.5 Sonnet按输入与输出token分别计费75% similar

Cite This Claim

Stable URI

https://kongchang.com/claim/584928

API

curl https://kongchang.com/api/v1/knowledge/claims/584928

MCP

get_claim(id=584928)