Unverified50% confidenceFactExact time
Meta's newly released 816B-parameter model activates only about 5% of its parameters per token, with the actual parameter count involved in computation on the order of 40B.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/21/2026
First Seen
Valid until: 1/19/2027
Sources
Related Entities
Related Claims
UnverifiedIn MoE models, each token is routed to only a small number of experts for computation, resulting in a large total parameter count but a small number of activated parameters.69% similarUnverifiedEach token in the GLM series 744B model activates only approximately 40B parameters, accounting for about 5.4% of the total.67% similarUnverifiedFor a configuration of 1.6 trillion total parameters and 49 billion activated parameters, only about 3% of parameters participate in computation during actual inference.66% similarUnverified高推理强度模式下单次请求的 Token 消耗可能是低强度模式的 5-20 倍65% similarVerifiedThe MoE architecture typically activates only 10%–20% of total parameters during each inference65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1083387API
curl https://kongchang.com/api/v1/knowledge/claims/1083387MCP
get_claim(id=1083387)