GLM-5.3 要上 LMArena 了,这次直接和前两代同台对比,想蹲 Z.ai 迭代幅度的话可以盯一下。
Z.ai 的 GLM-5.3 即将上线 LMArena 评测平台。官方将对比 GLM-5.1 与 GLM-5.2 的表现。此前 GLM-5.2 在 Agent Arena 的 5 项信号上取得进展,包括任务成功率与用户夸奖/抱怨等隐式反馈。新版本将检验这些改进能否延续。
GLM-5.3 by @Zai_org is coming to @arena soon! We’ll see how it compares to GLM-5.1 and GLM-5.2. GLM...
GLM-5.3 by @Zai_org is coming to @arena soon! We’ll see how it compares to GLM-5.1 and GLM-5.2. GLM's last update brought meaningful gains in the Agent Arena across both explicit human feedback (task success) and implicit human feedback (signals like praise and complaints). Arena.ai @arena Comparing GLM-5.2 (Max) with 5.1 visually across the 5 Agent Arena signals you can see where the newest model received the most gains: confirmed task successes and praise vs. complaint. 🔗 View Quoted Tweet 💬 1 🔄 0 ❤️ 43 👀 4925 📊 4 ⚡