Unverified50% confidenceBenchmarkExact time
In the MoE architecture, the Mixtral 8x7B model has a total of 46.7B parameters, but only approximately 12.9B parameters (2 experts) are activated per token, making inference costs close to those of a 13B dense model while achieving performance comparable to the 70B level.
1
Sources
50%
Confidence
Long-term
Relevance
9/6/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedMistral AI's Mixtral 8x7B adopts a MoE architecture with a total parameter count of approximately 47B, but only approximately 13B parameters are actually activated during each inference.77% similarUnverified在MoE模型中每个token只激活少量Expert（通常2-4个，Top-K路由），实际计算量可能仅相当于6-7B的Dense模型77% similarVerifiedMixtral 8x7B模型拥有8个专家，每个Token只激活2个，总参数量达到47B，但单次推理的计算量仅相当于一个13B参数的稠密模型76% similarUnverifiedMoE模型中'35B'指总参数量，每个token只激活少量Expert（通常2-4个），实际计算量可能仅相当于6-7B的Dense模型75% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/864754API
curl https://kongchang.com/api/v1/knowledge/claims/864754MCP
get_claim(id=864754)