Expired50% confidenceEventExact time
The research team consumed 11.7 billion (11.7B) tokens to evaluate the cybersecurity capabilities of multiple mainstream large language models.
1
Sources
50%
Confidence
Short-term (~14 days)
Relevance
9/3/2026
First Seen
Valid until: 9/17/2026(expired)
Sources
Related Entities
Related Claims
UnverifiedCommon Crawl是非营利组织运营的大规模网络爬取项目，每月抓取数十亿网页，数据集被用于训练包括GPT系列在内的大型语言模型66% similarUnverifiedThe context window of coding agents typically contains tens of thousands to hundreds of thousands of tokens of information.64% similarUnverified训练使用约500万行英文维基百科语料约3亿token，10万上下文词表，23758个名词目标64% similarUnverified深度研究单次任务消耗的token数量可达普通对话的数十倍，成本可能相当于数十次乃至上百次普通问答的资源消耗64% similarUnverifiedOpenAI动用Codex等模型分析了超过70亿条日志来调查该事件，耗费数百万GPU小时63% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/847670API
curl https://kongchang.com/api/v1/knowledge/claims/847670MCP
get_claim(id=847670)