Unverified50% confidenceBenchmarkExact time
Anthropic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%
1
Sources
50%
Confidence
Long-term
Relevance
7/20/2026
First Seen
Sources
GPT-5.6 & OpenAI Super App Deep Dive: Loop Engineering as the New Paradigm Replacing Prompt Engineering
bilibiliGoldenSpiderAI7/11/2026
Related Claims
UnverifiedAnthropic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）68% similarUnverifiedClaude第一次推出计算机使用能力时，OS World得分还不到30%66% similarUnverifiedIn a 2023 paper, Anthropic systematically measured sycophancy bias and found that even given obviously incorrect premises, a substantial proportion of responses from most mainstream models still chose to comply with the user.57% similarUnverifiedIn the OS World computer operation test, Sol surpassed Claude's Opus series while consuming 85% less compute.56% similarUnverifiedWhile launching Claude 4.5 Sonnet, Anthropic drastically cut the quota for Opus model usage by $200 Max plan subscribers, with usage limits dropping by approximately 20 times.56% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/566790API
curl https://kongchang.com/api/v1/knowledge/claims/566790MCP
get_claim(id=566790)