Dan Hendrycks releases multidimensional capability evaluation comparing Claude Opus 4.5 and Gemini 3
Original title:How does Claude Opus 4.5 compare to Gemini 3?
AI summary
AI safety researcher Dan Hendrycks published a cross-comparison of Claude Opus 4.5 and Google Gemini 3 on X, including index scores across text, vision and safety dimensions. Results show that after controlling for reasoning token counts, the two are comparable in text capability; Gemini 3 leads significantly on image/vision tasks; and Opus performs better in adversarial tests such as jailbreaks and honesty. The charts show text capability across 7 models (Gemini 3 Pro leading at 47.6), vision capability (Gemini 3 Pro highest at 57.1) and safety index (Opus lowest at 33.6, meaning safest).
Key Takeaways
- AI safety researcher Dan Hendrycks published a cross-comparison of Claude Opus 4.
- 5 and Google Gemini 3 on X,including index scores across text,vision and safety dimensions.
- Results show that after controlling for reasoning token counts,the two are comparable in text…
- Gemini 3 leads significantly on image/vision tasks
- and Opus performs better in adversarial tests such as jailbreaks and honesty.
Quick Facts
- Published
- Updated
- Source
- X:Dan Hendrycks (@hendrycks)
- Category
- Papers & Research
- AI score
- 78
该评测由知名 AI 安全研究者主导,数据维度全面(文本、视觉、安全),且直接对比当前头部模型,为读者提供了可量化的横向参考;尤其安全指标反常识(越低越好),有助于理解“能力”与“安全性”的权衡,对选型和风险评估有实用价值。
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: Dan Hendrycks releases multidimensional capability evaluation comparing Claude Opus 4.5 and Gemini 3. November 26, 2025. https://new.nat6.net/en/news/j5zfuprzkr3tehmqu76tdo4g4/