问HN:OpenAI模型是否有限制IP范围?
我观察到,OpenAI 模型的可用性似乎受到工具和/或 IP 范围的限制。有没有其他人也观察到类似的情况?
我并没有验证我的“组织”。所以我不能使用“llm -m o3”,这没问题。然而,在大学里,我可以使用 o3 和 o3-pro 以及 aider,没有任何错误。使用的是同一个 API 密钥。在家里,我无法使用 o3-pro 和 aider,错误信息大致是“chat/completion 不支持此模型,你是否使用了错误的端点?”不过,aider 与 o3 在家里是可以正常工作的。“llm -m o3”在家里无法使用。
我实际上检查了 aider 的源代码,想找出一些魔法,使得 o3 可用,而它与“llm -m o3”不兼容。但没有找到任何东西。这实在是有点奇怪。有没有其他人遇到类似的情况?
查看原文
I oserve that the availability of OpenAI models <i>seems</i> to be tool and/or IP-range restricted. Does anyone else observe something similar?<p>I have <i>NOT</i> verified my "organisation". So I can not use "llm -m o3", OK. However, at the university, I can use o3 and o3-pro with aider. No errors. Same API key. At home, I can not use o3-pro with aider, error is roughly "chat/completion does not support this model, did you use the wrong endpoint?". However, aider with o3 works at home. "llm -m o3" does not work at home.
I sacually checked the aider source code for some magic which would make o3 available while it does not work with "llm -m o3". Didn't find anything. This is rather strange. Does anyone else experience similar things?