Unverified80% confidenceFactTime unknown
使用Qwen3 27B Q4量化稠密模型实测，关闭思考模式下生成速度约58-60 Token/s
1
Sources
80%
Confidence
Medium-term (~90 days)
Relevance
6/2/2026
First Seen
Valid until: 8/31/2026
Sources
llama.cpp MTP加速部署指南：配置步骤与性能实测
bilibili尚书省说书人
Related Entities
Related Claims
UnverifiedQwen3.6 27B Q4量化版总分650分，排名第四74% similarUnverifiedQwen3 27B在256K长上下文记忆检索测试（大海捞针）中取得100%通过率，15次运行全部准确74% similarUnverified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73% similarUnverified原版Qwen3.6 27B Q4在工具调用测试中获得97分，说明蒸馏并不总能提升性能73% similarUnverifiedQwen3.6-27B在多个基准测试维度上全面超越了前代Qwen系列中参数量更大的旗舰模型72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/37327API
curl https://kongchang.com/api/v1/knowledge/claims/37327MCP
get_claim(id=37327)