Verified65% confidenceFactExact time

DeepSeek 系列模型采用 MoE 架构，每次推理时只激活总参数的10%-20%

3

Sources

65%

Confidence

Medium-term (~90 days)

Relevance

7/2/2026

First Seen

Valid until: 9/30/2026

Sources

十大AI编程模型终极评测：谁是代码天花板？

bilibiliEA同学6/8/2026

Related Claims

UnverifiedDeepSeek uses a MoE (Mixture of Experts) architecture, achieving efficient inference at large parameter scales through sparse activation79% similar UnverifiedDeepSeek的MoE架构使其计算成本仅为同等规模密集模型的十分之一左右78% similar

Cite This Claim

Stable URI

https://kongchang.com/claim/50322

API

curl https://kongchang.com/api/v1/knowledge/claims/50322

MCP

get_claim(id=50322)