AI Chat context got way worse in the last 2 days. Anyone else getting constant "Compressing conversation" even on small files?
Hey everyone, Genspark Plus user here. I use AI Chat daily with Claude Sonnet 4.6/Opus 4.6 mostly for research, project planning, outlining docs, that kind of stuff.
Up until like 2-3 days ago everything was fine. I could upload 10+ Word/PDF files, each around 50-100 pages, work through them in one chat session, no problems. The AI remembered everything, I could go deep into the material and keep asking follow-ups across multiple documents. Totally normal.
Then something changed. Now I upload ONE Word file, about 50 pages, and I immediately get "Compressing our conversation to conserve context, may take a few minutes..." Like right at the start, not after a long conversation. Just the first file and boom, compressing.
And then as I keep working it keeps hitting me with "For better performance, previous chat history has been compacted" over and over. After that the AI basically loses track of the document content. Gives vague answers, wrong details, can't recall stuff it literally just read. It's night and day compared to last week when the exact same workflow was smooth.
Nothing changed on my end. Same plan, same models, same file types, same workflow. So something definitely changed on Genspark's side.
I saw some older posts about this; like this one breaking down how Genspark compresses way before the model's actual context limit (https://www.reddit.com/r/genspark\_ai/comments/1tymiqy/why\_is\_there\_no\_information\_anywhere\_about\_chat/) and this one about limits getting progressively worse (https://www.reddit.com/r/genspark\_ai/comments/1tj9pwa/ai\_chat\_limit\_suggestion\_issues/). But what I'm experiencing now is way worse than what those posts described. Feels like the context window got cut significantly.
So Genspark team if you're reading this: did the context allocation for AI Chat get reduced recently? Is this intentional or a bug? Any plans to fix it?
Because honestly this is a dealbreaker for me. The whole point of using Genspark AI Chat for my work is handling large documents in long research sessions. If that's gone I need to know so I can move on.
Anyone else noticing the same thing in the last few days? Would love to know if it's just me or widespread.