AI harness affects performance more than the model itself: GPT-5.5 and Opus 4.7 score higher in third-party tools
Original title:Aftermarket Harnesses
AI summary
Endor Labs testing shows the same model performs very differently across harnesses: GPT-5.5 achieved 87.2% functional correctness in Cursor, far above its 61.5% in the native Codex environment; Opus 4.7 scored 91.1% in Cursor versus 87.2% in Claude Code. The harness determines what context is sent and the caching strategy, and smart caching can save 40-80% in costs and speed things up by 13-31%.
Key Takeaways
- Endor Labs testing shows the same model performs very differently across harnesses: GPT-5.
- 5 achieved 87.
- 2% functional correctness in Cursor,far above its 61.
- 5% in the native Codex environment
- 7 scored 91.
Quick Facts
- Published
- Updated
- Source
- Tomasz Tunguz 博客(VC 分析)
- Category
- Opinion
- AI score
- 67
作者引用同一模型在不同 harness 下的分数差与缓存研究,说明 harness 而非模型决定成本和表现,读者可借此审视自己的工具链选择。
Related entities
Source & copyright
This article is based on curated titles and AI summaries from the AIHOT public API, with attribution and links back. Original copyright remains with the original publishers and authors; this site does not mirror full texts.
How to Cite
When citing this article, keep the title, date, and permalink, and credit Orange News as the source.
Orange News: AI harness affects performance more than the model itself: GPT-5.5 and Opus 4.7 score higher in third-party tools. July 28, 2026. https://new.nat6.net/en/news/cmtggvk4o0chbroj7ptn943ah/