gpt-imageIMAGE
GPT Image 2.5 moderation tightens, downgrading the model works around it
A user reports that GPT Image 2.5 moderation has become stricter, with prompts that previously worked now failing, but downgrading the model lets images generate normally again.
李岳
@liyue_aiArchived 2026.09.14Originally posted 2026.09.12
李岳 @liyue_aiOriginally posted 2026.09.12
最近使用GPT Image 2.5发现审查力度进一步加大， 以前可行的一些提示词，突然就会收到出图失败提示。 但发现将模型降级之后，就可以正常出图了。 有遇到类似情况的不妨试一下。
x.com/liyue_ai/status/2098670001637200173Archived 2026.09.14 19:39