Compare latest models: https://artificialanalysis.ai/
GLM 5.2 was very important because it was the fastest open source model.
If you are reading this... I'm BEGGING YOU: Take the weights.
It will cost you very little (1 TB HDD for 5.2) and might benefit you very much one day.
At least get Deepseek Flash, the cheapest model to run, for < 200GB! You can go lighter still with quantized models.
Please ignore my last messages if you saw them, I went to the wrong site lol. disinformation.