Unverified50% confidenceFactExact time

Step 5 的模型体量远小于同类旗舰竞品

1

Sources

50%

Confidence

Medium-term (~90 days)

Relevance

9/24/2026

First Seen

Valid until: 12/23/2026

Sources

阶跃星辰Step 5预览版实测：小模型如何逼近Kimi与GLM

Related Entities

Related Claims

Unverified在需要5步以上操作的任务中，最强模型的成功率会从首步的80%以上骤降至后续步骤的40%甚至更低，呈现指数级下降特征68% similar Unverified研究额外引入六个被刻意削弱、能力明显更差的模型作为正向对照67% similar Unverified一个早期迭代版本在IFEval指令遵循基准测试上比Huihui母模型低了五个百分点67% similar Unverified在MATH数据集的Level 5竞赛级难题上，前沿模型准确率仍低于60%66% similar UnverifiedGLM5.2的推理速度明显慢于同级别模型，是其最大短板66% similar

Cite This Claim

Stable URI

https://kongchang.com/claim/947144

API

curl https://kongchang.com/api/v1/knowledge/claims/947144

MCP

get_claim(id=947144)