Unverified50% confidenceOpinionExact time

据社区消息，通义千问新版本目标是让能力超越GLM 5.2的水平

1

Sources

50%

Confidence

Medium-term (~90 days)

Relevance

7/22/2026

First Seen

Valid until: 10/20/2026

Sources

通义千问、DeepSeek V4、GLM新版同期曝光：国产大模型迭代加速

Related Claims

UnverifiedGLM 5.2 is described as a new benchmark for open-weight models70% similar UnverifiedGLM 5.2支持100万Token上下文，经实测真实可用70% similar UnverifiedGLM 5.2引入了多令牌预测（MTP）技术，该技术由Meta在2024年研究中提出，可通过投机解码提升推理速度69% similar

Cite This Claim

Stable URI

https://kongchang.com/claim/589978

API

curl https://kongchang.com/api/v1/knowledge/claims/589978

MCP

get_claim(id=589978)