gpt-6POST
User complains Astra xhigh seems downgraded
A Chinese user says GPT-6 Astra xhigh produced a cost-comparison plan that was still unusable after several revisions, while Opus 5.5 read the codex session and produced a usable result in one shot.
supermao
@buaaxhmArchived 2026.10.01Originally posted 2026.09.30
supermao @buaaxhmOriginally posted 2026.09.30
对 o/一点好感都没有了 我用 GPT-6 Astra xhigh 让它出一个成本对比核算方案 也不知道它挂羊头卖狗肉降到了什么模型，在那迭代好几版还是一坨屎 然后给 Opus 5.5直接读 codex session，OneShot 直出，清清爽爽，干干净净，只需要小改就能用 你们就这样玩你们的用户吧，负分，滚粗 @ChatGPT @sama @thsottiaux
x.com/buaaxhm/status/2105103220351930701Archived 2026.10.01 10:10