Unverified60% confidenceFactExact time

早期GPT-4在原始SWE-Bench上的通过率仅约1.7%

2

Sources

60%

Confidence

Medium-term (~90 days)

Relevance

7/2/2026

First Seen

Valid until: 9/30/2026

Sources

GLM4开源旗舰实测：前端开发碾压Claude等商业模型

bilibiliAI-seeker6/21/2026

Related Claims

Unverified早期GPT-4在SWE-bench上的解决率不足2%，而近期顶尖模型已突破50%82% similar UnverifiedGPT-4时代的上下文限制为8192个token76% similar UnverifiedGPT-5.6在Terminal-Bench 2.1编码榜单上以Ultra配置取得91.9%的成绩76% similar VerifiedGPT-5.4在SWE-Bench Pro测试中拿下57.7%准确率，超越了GPT-5.3-Codex的56.8%74% similar

Cite This Claim

Stable URI

https://kongchang.com/claim/40631

API

curl https://kongchang.com/api/v1/knowledge/claims/40631

MCP

get_claim(id=40631)