Unhappy with GLM 5.3 Flash now
For a long time I was very happy with GLM 5.3 Flash. It worked its way through large codebases, and the differences from GLM 5.3 were barely noticeable. Only very rarely did GLM 5.3 Flash get stuck, and when it did, those were genuinely very complex issues that GLM 5.3 then solved. But for the past few days, almost a week now, the model has been giving me serious trouble. Careless mistakes keep happening. Things that were discussed in the same session get forgotten. It takes 5, 6, 7 attempts to fix bugs, and even then only because I keep steering it back on track. Right now it is completely unsuitable for autonomous work. This shows up especially after 1:00 a.m. (UTC+2). GLM 5.3 has also gotten a bit worse, but not nearly as badly. I experience also lower speed than in the good old days. Have you had similar experiences?