Unverified50% confidenceFactExact time
Opus 4.7每次解题的中位输出token数为60,000，而GPT 5.5仅需16,000
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
7/2/2026
First Seen
Valid until: 9/30/2026
Sources
DeepSWE基准测试揭示真相：GPT 5.5大幅领先Opus 4.7
bilibili7k的每日搬运5/29/2026
Related Claims
Unverified早期GPT-4在SWE-bench的通过率仅为1.7%，Claude 3 Opus约为4.8%，直到2024年中才突破20%79% similarPartially VerifiedClaude Opus 4.8在SWE-bench基准测试中得分69.2%，GPT 5.5为58.6%，Google Gemini为54.2%79% similarUnverifiedOpus 4.7在长上下文（50万-100万token）检索方面表现明显退步，远不如DeepSeek和GPT 5.579% similarUnverified在Agents' Last Exam评测中，GPT-5.6 Sol取得53.6分，比采用自适应推理的Claude Fable 5高出13.1分75% similarUnverifiedClaude Opus 5在智能体搜索BrowseComp上得分30.8%，与GPT系列基本持平；工具增强多学科推理64.7%；生物学人类可解问题上高达90.1%75% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/47588API
curl https://kongchang.com/api/v1/knowledge/claims/47588MCP
get_claim(id=47588)