Unverified50% confidenceBenchmarkExact time
GLM-5.2 采用 IndexShare 技术，每四个稀疏注意力层复用同一个索引器，在 1M 上下文长度下将每 token 的 FLOPs 降低了 2.9 倍
1
Sources
50%
Confidence
Long-term
Relevance
8/27/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedFan's team's performance upper-bound experiment found that under ideal conditions, computational FLOPs can be reduced by 75×, while accuracy improves by 1.1%.60% similarUnverified金融风控中可以使用F2分数(β=2)来更强调查全率，垃圾邮件过滤中可以使用F0.5分数(β=0.5)来更强调查准率59% similarUnverifiedGLM 5.2's inference speed is approximately 106 tokens per second.58% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/807813API
curl https://kongchang.com/api/v1/knowledge/claims/807813MCP
get_claim(id=807813)