ChatGPT Go users: Has Luna caused a noticeable drop in image understanding and context handling?
retained text
ChatGPT Go: Has image understanding and context handling gotten worse since the Luna split? I'm a ChatGPT Go user, and since the recent Luna split, I've been feeling a noticeable drop in response quality. I know Go has usage limits, and I don't have a problem with having limits on usage. What I don't understand is why a paid plan can feel less capable at basic context and image understanding tasks than it did before. Today I was able to reproduce a particularly strange example. I used an adult manga only as a test case for image understanding. I am NOT posting the images here because this is a SFW subreddit. The story involved a husband waiting for his wife to return. Her return date happened to be her birthday, so he had prepared a birthday celebration for her. However, the wife does not return. Instead, the husband is informed of what happened through a video call from the person who took her away. The story ends with the husband being left behind. I showed ChatGPT the final page and asked about the meaning of the ending. ChatGPT interpreted it as a happy ending — essentially reading the "Welcome home!" and "Happy Birthday!" decorations as evidence that the wife had returned. But she had NOT returned. Those decorations had been prepared in anticipation of her return. The point of the final page was that the preparations remained while she never came back. What makes this interesting is that I was able to reproduce the same misinterpretation even when I showed only the final page separately, rather than making it process the entire sequence of images. So this doesn't seem to be simply a case of an extremely long conversation overwhelming the context. I've also encountered other problems around context and continuity. In a separate case, a Project was not properly incorporating information from the preceding conversation/material. That issue has already been reported to OpenAI Support as Case 14173404, and Support requested a screen recording showing the problem. I've also previously experienced a mobile ChatGPT issue where copying text from a conversation can select/pick up a huge amount of additional text instead of the intended section. That problem appeared to recur today as well. So I'm seeing several things that make me wonder whether something changed around the Luna split: \- poorer image/story interpretation \- failure to connect visual information with conversational context \- weaker continuity/context handling \- Project context/reference problems \- mobile copy/paste selecting far more content than intended I'm not claiming that all of these necessarily have the same technical cause. I'm also not claiming that Luna itself is definitely responsible. But the timing is noticeable to me: these problems have become much more apparent since the Luna split. For Go users, I think this is an important distinction. I'm perfectly willing to accept usage limits as part of a lower-cost paid plan. But if the trade-off is supposed to be limits on usage rather than a reduction in basic reasoning, image understanding, or context handling, then it's reasonable to ask what exactly changed. Has anyone else using Go noticed similar problems with Luna, particularly with image understanding, context retention, or Projects? Again, I'm not posting the adult images themselves. The adult manga was simply the test material used to reproduce the image-understanding problem.
document fetched 2026-09-03 00:21 · canonical https://reddit.com/r/ChatGPT/comments/1w5c54s/chatgpt_go_users_has_luna_caused_a_noticeable