Expired60% confidenceFactExact time
In the MoE architecture, a 600-billion-parameter model might only activate 37 billion parameters per inference
2
Sources
60%
Confidence
Medium-term (~90 days)
Relevance
7/2/2026
First Seen
Valid until: 9/30/2026(expired)
Sources
DeepSeek V4 Pro In-Depth Review: Performance Rivaling GPT-5.5 at 1/12 the Cost
bilibili63号炼金工坊6/18/2026
Related Claims
UnverifiedMoE架构下模型总参数量远大于单次推理实际激活的参数量，Qwen3.8每次推理激活参数可能仅数百亿级别79% similarUnverifiedThe MoE architecture activates only a small subset of expert sub-networks during inference; a 2.4T-parameter model may activate only approximately 20–40 billion parameters per inference.79% similarUnverifiedThe MoE architecture has fewer than one-tenth of its parameters actually participating in computation at any given moment, allowing the model to maintain the knowledge capacity of a large model while retaining the inference speed and computational cost of a small model.75% similarVerified1.6万亿参数级别的巨型模型通常采用混合专家架构（MoE），每次推理只激活一小部分专家子网络71% similarUnverifiedAlthough the MoE architecture activates fewer parameters, all 671B parameters still need to be loaded into VRAM for on-demand use, and the model's VRAM consumption is not significantly reduced by the MoE architecture.71% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/44333API
curl https://kongchang.com/api/v1/knowledge/claims/44333MCP
get_claim(id=44333)