Union Alpha 首测:像多模型编排、极慢、爱卡循环
用户实测 Union Alpha 后认为它不像任何中国模型:简单任务也调子代理、行为像 Sakana Fugu、极慢且陷入死循环,一次测试跑 2 小时最后说做不完,怀疑是新实验室把多个模型编排成一个。

first impression of union alpha this doesn’t feel like any of the chinese models i’ve tested, so far it: > calls subagents even for simple tasks > behaves a lot like sakana fugu model > super slow > keeps getting stuck in loops without doing anything one of my tests ran for 2 hours just to come back and say it couldn’t complete the task 😭 i think this model is from a new lab orchestrating multiple models into one new model i could be completely wrong, but few other people testing it noticed same behavior too chinese labs already have good compute and their models usually don’t behave like this who do you think is behind union alpha?