Unverified50% confidenceFactExact time

Qwen3.8-27B默认开启Thinking模式，推理深度分为Low/Medium/X-High三档

1

Sources

50%

Confidence

Long-term

Relevance

9/3/2026

First Seen

Sources

Qwen3.8-27B本地部署实测：27B参数真能媲美Claude Opus？

Related Entities

Qwen3.8-27Bmodel

Related Claims

Partially VerifiedQwen 3 Max和DeepSeek-V3均采用MoE架构，总参数量达数千亿，但实际推理激活参数仅为总量的一小部分74% similar UnverifiedMAI Thinking-E在数学、科学等推理密集型领域表现较强，但在编码测试中仅达到DeepSeek V3.2的水平71% similar UnverifiedQwen3-4B在多项推理类评测指标上已与DeepSeek V3或GPT-4o相差无几70% similar UnverifiedQwen3 Max在生成精细度方面与Claude Opus级别及更高阶闭源模型存在可见差距70% similar

Cite This Claim

Stable URI

https://kongchang.com/claim/848868

API

curl https://kongchang.com/api/v1/knowledge/claims/848868

MCP

get_claim(id=848868)