X：ARC Prize (@arcprize)· X：ARC Prize (@arcprize)·· 6 天前AI 评分48
GLM 5.3 Flash 在 ARC-AGI 基准测试成绩提升
AI 导读
@Zai_org 当 GLM 5.3 Flash 以 Ox Alpha 的代号在 @OpenRouter 上 stealth 上线时，我们在 ARC-AGI-1 和 2 的公开任务上进行了测试。后来我们通过 @Baseten 测试了官方版 GLM 5.3 Flash，发现两项基准测试的成绩平均提升了 15 个百分点。
正文
@Zai_org While GLM 5.3 Flash was available in stealth as Ox Alpha on @OpenRouter, we tested it on ARC-AGI-1 & 2 public tasks. When we later tested the official GLM 5.3 Flash through @Baseten, we saw scores rise by an average of 15 percentage points across both benchmarks.
来源：X：ARC Prize (@arcprize) · x.com