AZ

[AZ @Pyqx369]

@pumpingecum @Gh0ul_07 @teortaxesTex I'm not sure if this qualifies as an example, but to prove its reverse‑engineering ability, one of our people used it to create a CS hack. We haven't tested the same prompt on Fable, so we can't say for certain whether it would set off Fable's safety filters

[AZ @Pyqx369]

@Gh0ul_07 @teortaxesTex We have also provided responses to several biochemistry questions, which Fable or Opus are unable to handle

[AZ @Pyqx369]

@Gh0ul_07 @teortaxesTex Yes, you're about to see even more astonishing examples. Everyone who's tried it is blown away. I don't know whether Deepseek has put up restrictions, but currently it seems that only Chinese IPs can use it – and even that is hit-or-miss.

[AZ @Pyqx369]

@teortaxesTex Your information is way too outdated. You probably have no idea how extensive today's gray‑scale testing is. Just sit tight and wait until the news reaches you

[AZ @Pyqx369]

@2youg1 @teortaxesTex 搁这搞笑呢，真以为大家没用过fable一样，百亿补贴那天谁没玩过fable，根本不是一个级别的模型，现在灰度还在继续，自己试试就知道了

[AZ @Pyqx369]

@MaxForAI 你这引流狗别讲着讲那了，还AI从业人员，你有个屁的内部消息，昨天吹今天骂明天又继续吹

[AZ @Pyqx369]

@teortaxesTex Why don't you try Fable with the same prompts? We can provide all the prompts and session logs, rather than using CoT to evaluate

[AZ @Pyqx369]

@teortaxesTex Also, here is the website where we are collecting the products for this test:https://t.co/zXYUCD0gt3

[AZ @Pyqx369]

@tianyi 今天这模型发的是不是有问题，和灰测的差距不是一星半点？甚至只是比flash略强？

[AZ @Pyqx369]

@tianyi 崔啊，这都8月10号了，harness和PRO呢，总得端上来一个吧，继七月中旬之后又要重新定义八月初吗，怎么审内测资格都要审一个星期，那测还得测多久？做不到就不要写官网上去给人期待啊