Verified65% confidenceFactTime unknown
Qwen3.6 35B-A3B是混合专家架构（MoE），总参数量35B，每次推理实际激活参数量约3B
3
Sources
65%
Confidence
Long-term
Relevance
6/1/2026
First Seen
Sources
Qwen3.6 MTP加速实测：单GPU推理飙到220 token/s
bilibili普通鱼学家呀
Related Entities
Cite This Claim
Stable URI
https://kongchang.com/claim/18060API
curl https://kongchang.com/api/v1/knowledge/claims/18060MCP
get_claim(id=18060)