Unverified50% confidenceFactExact time
DeepSeek-R1-Distill-Qwen-32B 以 Qwen-2.5-32B 为基座，通过从完整的 671B MoE 架构模型中蒸馏知识而来
1
Sources
50%
Confidence
Long-term
Relevance
9/10/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedQwen3.6-27B采用稠密（Dense）架构，而非MoE（Mixture of Experts）架构74% similarPartially VerifiedQwen3.8-Flash-Next 采用 125B 总参数、6B 激活参数的混合专家（MoE）架构72% similarUnverifiedMoE架构的典型代表包括Mixtral 8x7B和DeepSeek-V272% similarUnverifiedAdaptiveSpec支持DeepSeek-R1-Distill-Llama-8B、Llama-3.1-8B-Instruct和Qwen3-8B三个主流目标模型71% similarVerifiedQwen3 Max 采用 MoE（混合专家）架构70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/888304API
curl https://kongchang.com/api/v1/knowledge/claims/888304MCP
get_claim(id=888304)