Unverified85% confidenceFactTime unknown
Anthropic的Computer Use在OSWorld基准测试上的操作准确率约为14.9%（2024年10月数据）
1
Sources
85%
Confidence
Long-term
Relevance
6/1/2026
First Seen
Sources
computer-use-windows：在Windows上构建AI桌面代理的开源框架
githubCam10001110101
Related Entities
Related Claims
UnverifiedAnthropic于2024年10月率先发布了Claude的Computer Use功能，其论文披露在OSWorld基准上成功率约22%，显著低于人类的72%68% similarUnverifiedAccording to the OSWorld benchmark, the most advanced Computer Use models still achieve less than 40% task success rate64% similarUnverified在网页浏览基准BrowseComp上Sol拿到92.2%，在OSWorld 2.0上拿到62.6%，两项均为当前最佳成绩，OSWorld上超过Opus 4.8的同时输出Token减少85%62% similarUnverified在OS World电脑操作测试中，Sol超过了Claude的Opus系列，消耗算力少85%59% similarUnverifiedVirtioFS在元数据密集型操作上相比osxfs部分基准测试显示性能提升可达20倍以上58% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/29286API
curl https://kongchang.com/api/v1/knowledge/claims/29286MCP
get_claim(id=29286)