Unverified85% confidenceFactTime unknown

BugFind测试中90分以上的仅有3个模型，表现最好的是APEX量化的35B MOE

1

Sources

85%

Confidence

Medium-term (~90 days)

Relevance

6/1/2026

First Seen

Valid until: 8/30/2026

Sources

Qwen3.6量化版本地部署实测：NVFP4、APEX、Q4、Q6哪个最值得选

bilibili尚书省说书人

Related Entities

Qwen3.6 35B MOE APEXmodel

Related Claims

Unverified在BugFind代码找Bug测试中，Qwen3.6 27B Q4精度得了87分，比Q6的80分更高，出现了量化反转现象63% similar Unverified混元3是混合专家模型（MoE），总参数量为295B，激活参数约21B62% similar Unverified千问3的235B参数MoE旗舰版在多项权威基准测试中接近甚至超越闭源顶级模型62% similar UnverifiedMuse-Glimmer-30B在测试中通过了88%的题目61% similar Unverified千问3.6的35B MOE模型每次只激活其中3B的参数61% similar

Cite This Claim

Stable URI

https://kongchang.com/claim/17967

API

curl https://kongchang.com/api/v1/knowledge/claims/17967

MCP

get_claim(id=17967)