Bug report

Hi, I think as with many of the past feedback and reports already raised - response cutoff is a very frequent bug that happens across very many models. Oftentimes, when given a prompt, and if the response seems like it would be long, the model produces either a response that is cutoff halfway (or even earlier), or worse, produces no response at all. Please look into this and fix, please.

The GPT 5 Pro has been consistently very slow and has not responded effectively to my queries. This is such a hassle. I went directly to CHatgpt, used their native version of GPT 5 Pro and it worked so fast and did what I asked it to do. I don't know what limitations you guys have set, but this is really annoying. I find problems with all the expensive models. No a happy customer.

I've already had several conversation flows where GPT 5 forgot all or at least some of the messages in the middle of the conversation and suddenly asked what it was all about. I don't know if the problem is with Merlin or with OpenAI, but if it comes from Merlin, it should be fixed. In the attached image, for example, I was suddenly asked which product category it was about, although GPT 5 only recommended products in the desired category (in this case amplifier) to me just one message above

Hi, I’m using Claude 4 Sonnet (Thinking), and I noticed a strange issue. When I ask a question, Claude gives a full answer but then it starts “Thinking...” again right after, and gives me the same answer a second time. I’m not clicking anything, it just happens automatically. This repeats the whole reply and uses up more tokens. It happens even when I just send one prompt, and it keeps happening every time.

prompt: now create one for the physics of solar masses, make it so the mass can be a black hole or a start or a planet. Be creative. Make it like the last one where it is interactive and students can learn from it. The code snippet that it was creating stopped suddenly and said that the craft was done.

I have created a bot and added two PDFs to the knowledge base, but the bot does not seem to recognise that there is anything there.

And never finishes, gets an error. Please fix this.

This happening a lot more, especially with Gemini 2.5 Pro. I have wasted so many credits asking for the model to complete its response. This is especially common where it says a Craft was created and wasn't, so I keep asking for it to repeat and provide. This wastes a lot of credits for me.