Unverified50% confidenceOpinionExact time
分析推测该模型是2到4万亿参数规模的MoE模型，每个Token仅激活约1500亿参数
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
7/8/2026
First Seen
Valid until: 10/6/2026
Sources
GPT旗舰模型集成Codex：750 Token/秒推理速度详解
bilibili声译看世界7/7/2026
Related Claims
Unverified1000亿参数的模型处理单次推理请求需约80GB显存，百亿参数级别模型仅需约20GB77% similarUnverified若MoE模型每次推理仅激活约1/8的参数，实际计算开销可能仅相当于百亿参数级别的稠密模型75% similarUnverified一个千亿参数级别旗舰模型每次推断所需GPU资源可能比百亿参数模型高出10至20倍74% similarUnverifiedTiny-MoE模型总参数量约200M，但每个token激活的参数仅约90M74% similarUnverified对于 198B 总参数的稀疏 MoE 模型，实际激活参数量可能仅为 20-40B 量级73% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/291491API
curl https://kongchang.com/api/v1/knowledge/claims/291491MCP
get_claim(id=291491)